diff --git "a/trainer_log.jsonl" "b/trainer_log.jsonl" --- "a/trainer_log.jsonl" +++ "b/trainer_log.jsonl" @@ -7653,3 +7653,429 @@ {"current_steps": 38178, "total_steps": 42420, "eval_loss": 0.15290038287639618, "epoch": 18.0, "percentage": 90.0, "elapsed_time": "2:18:18", "remaining_time": "0:15:22", "throughput": 4628.33, "total_tokens": 38408800} {"current_steps": 38180, "total_steps": 42420, "loss": 0.1137, "lr": 1.50698095574002e-06, "epoch": 18.000942951438002, "percentage": 90.0, "elapsed_time": "2:18:20", "remaining_time": "0:15:21", "throughput": 4627.48, "total_tokens": 38410912} {"current_steps": 38185, "total_steps": 42420, "loss": 0.0959, "lr": 1.5034657189554824e-06, "epoch": 18.003300330033003, "percentage": 90.02, "elapsed_time": "2:18:22", "remaining_time": "0:15:20", "throughput": 4627.52, "total_tokens": 38418112} +{"current_steps": 38190, "total_steps": 42420, "loss": 0.0534, "lr": 1.4999544597352505e-06, "epoch": 18.005657708628007, "percentage": 90.03, "elapsed_time": "2:18:23", "remaining_time": "0:15:19", "throughput": 4627.52, "total_tokens": 38423040} +{"current_steps": 38195, "total_steps": 42420, "loss": 0.0248, "lr": 1.4964471786737184e-06, "epoch": 18.008015087223008, "percentage": 90.04, "elapsed_time": "2:18:24", "remaining_time": "0:15:18", "throughput": 4627.53, "total_tokens": 38427520} +{"current_steps": 38200, "total_steps": 42420, "loss": 0.0336, "lr": 1.492943876364611e-06, "epoch": 18.010372465818012, "percentage": 90.05, "elapsed_time": "2:18:25", "remaining_time": "0:15:17", "throughput": 4627.54, "total_tokens": 38432032} +{"current_steps": 38205, "total_steps": 42420, "loss": 0.099, "lr": 1.4894445534009816e-06, "epoch": 18.012729844413013, "percentage": 90.06, "elapsed_time": "2:18:26", "remaining_time": "0:15:16", "throughput": 4627.57, "total_tokens": 38437504} +{"current_steps": 38210, "total_steps": 42420, "loss": 0.085, "lr": 1.4859492103752037e-06, "epoch": 18.015087223008017, "percentage": 90.08, "elapsed_time": "2:18:27", "remaining_time": "0:15:15", "throughput": 4627.6, "total_tokens": 38442240} +{"current_steps": 38215, "total_steps": 42420, "loss": 0.0383, "lr": 1.4824578478789763e-06, "epoch": 18.017444601603017, "percentage": 90.09, "elapsed_time": "2:18:28", "remaining_time": "0:15:14", "throughput": 4627.62, "total_tokens": 38447232} +{"current_steps": 38220, "total_steps": 42420, "loss": 0.1113, "lr": 1.4789704665033321e-06, "epoch": 18.019801980198018, "percentage": 90.1, "elapsed_time": "2:18:29", "remaining_time": "0:15:13", "throughput": 4627.63, "total_tokens": 38451936} +{"current_steps": 38225, "total_steps": 42420, "loss": 0.0824, "lr": 1.475487066838621e-06, "epoch": 18.022159358793022, "percentage": 90.11, "elapsed_time": "2:18:30", "remaining_time": "0:15:12", "throughput": 4627.62, "total_tokens": 38456896} +{"current_steps": 38230, "total_steps": 42420, "loss": 0.2471, "lr": 1.4720076494745243e-06, "epoch": 18.024516737388023, "percentage": 90.12, "elapsed_time": "2:18:31", "remaining_time": "0:15:10", "throughput": 4627.64, "total_tokens": 38461568} +{"current_steps": 38235, "total_steps": 42420, "loss": 0.1113, "lr": 1.468532215000054e-06, "epoch": 18.026874115983027, "percentage": 90.13, "elapsed_time": "2:18:32", "remaining_time": "0:15:09", "throughput": 4627.65, "total_tokens": 38466080} +{"current_steps": 38240, "total_steps": 42420, "loss": 0.1319, "lr": 1.4650607640035342e-06, "epoch": 18.029231494578028, "percentage": 90.15, "elapsed_time": "2:18:33", "remaining_time": "0:15:08", "throughput": 4627.67, "total_tokens": 38471296} +{"current_steps": 38245, "total_steps": 42420, "loss": 0.0225, "lr": 1.4615932970726276e-06, "epoch": 18.031588873173032, "percentage": 90.16, "elapsed_time": "2:18:34", "remaining_time": "0:15:07", "throughput": 4627.7, "total_tokens": 38476608} +{"current_steps": 38250, "total_steps": 42420, "loss": 0.2479, "lr": 1.4581298147943179e-06, "epoch": 18.033946251768032, "percentage": 90.17, "elapsed_time": "2:18:35", "remaining_time": "0:15:06", "throughput": 4627.73, "total_tokens": 38481888} +{"current_steps": 38255, "total_steps": 42420, "loss": 0.052, "lr": 1.454670317754911e-06, "epoch": 18.036303630363037, "percentage": 90.18, "elapsed_time": "2:18:36", "remaining_time": "0:15:05", "throughput": 4627.72, "total_tokens": 38486560} +{"current_steps": 38260, "total_steps": 42420, "loss": 0.0541, "lr": 1.4512148065400465e-06, "epoch": 18.038661008958037, "percentage": 90.19, "elapsed_time": "2:18:37", "remaining_time": "0:15:04", "throughput": 4627.74, "total_tokens": 38491456} +{"current_steps": 38265, "total_steps": 42420, "loss": 0.1157, "lr": 1.4477632817346764e-06, "epoch": 18.04101838755304, "percentage": 90.21, "elapsed_time": "2:18:38", "remaining_time": "0:15:03", "throughput": 4627.78, "total_tokens": 38496992} +{"current_steps": 38270, "total_steps": 42420, "loss": 0.3084, "lr": 1.4443157439230858e-06, "epoch": 18.043375766148042, "percentage": 90.22, "elapsed_time": "2:18:39", "remaining_time": "0:15:02", "throughput": 4627.77, "total_tokens": 38501728} +{"current_steps": 38275, "total_steps": 42420, "loss": 0.0447, "lr": 1.4408721936888913e-06, "epoch": 18.045733144743046, "percentage": 90.23, "elapsed_time": "2:18:40", "remaining_time": "0:15:01", "throughput": 4627.79, "total_tokens": 38506272} +{"current_steps": 38280, "total_steps": 42420, "loss": 0.0947, "lr": 1.4374326316150182e-06, "epoch": 18.048090523338047, "percentage": 90.24, "elapsed_time": "2:18:41", "remaining_time": "0:15:00", "throughput": 4627.81, "total_tokens": 38512000} +{"current_steps": 38285, "total_steps": 42420, "loss": 0.1303, "lr": 1.4339970582837343e-06, "epoch": 18.05044790193305, "percentage": 90.25, "elapsed_time": "2:18:42", "remaining_time": "0:14:58", "throughput": 4627.85, "total_tokens": 38517344} +{"current_steps": 38290, "total_steps": 42420, "loss": 0.1124, "lr": 1.430565474276621e-06, "epoch": 18.05280528052805, "percentage": 90.26, "elapsed_time": "2:18:43", "remaining_time": "0:14:57", "throughput": 4627.85, "total_tokens": 38521440} +{"current_steps": 38295, "total_steps": 42420, "loss": 0.1681, "lr": 1.4271378801745867e-06, "epoch": 18.055162659123056, "percentage": 90.28, "elapsed_time": "2:18:44", "remaining_time": "0:14:56", "throughput": 4627.82, "total_tokens": 38525568} +{"current_steps": 38300, "total_steps": 42420, "loss": 0.0516, "lr": 1.4237142765578664e-06, "epoch": 18.057520037718056, "percentage": 90.29, "elapsed_time": "2:18:45", "remaining_time": "0:14:55", "throughput": 4627.85, "total_tokens": 38530336} +{"current_steps": 38305, "total_steps": 42420, "loss": 0.222, "lr": 1.4202946640060138e-06, "epoch": 18.05987741631306, "percentage": 90.3, "elapsed_time": "2:18:46", "remaining_time": "0:14:54", "throughput": 4627.87, "total_tokens": 38535264} +{"current_steps": 38310, "total_steps": 42420, "loss": 0.0651, "lr": 1.4168790430979128e-06, "epoch": 18.06223479490806, "percentage": 90.31, "elapsed_time": "2:18:47", "remaining_time": "0:14:53", "throughput": 4627.89, "total_tokens": 38540064} +{"current_steps": 38315, "total_steps": 42420, "loss": 0.0242, "lr": 1.4134674144117733e-06, "epoch": 18.064592173503065, "percentage": 90.32, "elapsed_time": "2:18:49", "remaining_time": "0:14:52", "throughput": 4627.94, "total_tokens": 38546976} +{"current_steps": 38320, "total_steps": 42420, "loss": 0.063, "lr": 1.4100597785251217e-06, "epoch": 18.066949552098066, "percentage": 90.33, "elapsed_time": "2:18:50", "remaining_time": "0:14:51", "throughput": 4627.98, "total_tokens": 38552128} +{"current_steps": 38325, "total_steps": 42420, "loss": 0.0469, "lr": 1.4066561360148167e-06, "epoch": 18.06930693069307, "percentage": 90.35, "elapsed_time": "2:18:51", "remaining_time": "0:14:50", "throughput": 4627.99, "total_tokens": 38557088} +{"current_steps": 38330, "total_steps": 42420, "loss": 0.0322, "lr": 1.4032564874570359e-06, "epoch": 18.07166430928807, "percentage": 90.36, "elapsed_time": "2:18:52", "remaining_time": "0:14:49", "throughput": 4628.0, "total_tokens": 38562112} +{"current_steps": 38335, "total_steps": 42420, "loss": 0.1247, "lr": 1.3998608334272829e-06, "epoch": 18.074021687883075, "percentage": 90.37, "elapsed_time": "2:18:53", "remaining_time": "0:14:47", "throughput": 4627.97, "total_tokens": 38566112} +{"current_steps": 38340, "total_steps": 42420, "loss": 0.0259, "lr": 1.3964691745003867e-06, "epoch": 18.076379066478076, "percentage": 90.38, "elapsed_time": "2:18:54", "remaining_time": "0:14:46", "throughput": 4628.0, "total_tokens": 38572000} +{"current_steps": 38345, "total_steps": 42420, "loss": 0.2143, "lr": 1.393081511250488e-06, "epoch": 18.07873644507308, "percentage": 90.39, "elapsed_time": "2:18:55", "remaining_time": "0:14:45", "throughput": 4628.01, "total_tokens": 38577024} +{"current_steps": 38350, "total_steps": 42420, "loss": 0.0201, "lr": 1.3896978442510672e-06, "epoch": 18.08109382366808, "percentage": 90.41, "elapsed_time": "2:18:56", "remaining_time": "0:14:44", "throughput": 4628.02, "total_tokens": 38581152} +{"current_steps": 38355, "total_steps": 42420, "loss": 0.1473, "lr": 1.3863181740749243e-06, "epoch": 18.083451202263085, "percentage": 90.42, "elapsed_time": "2:18:57", "remaining_time": "0:14:43", "throughput": 4628.04, "total_tokens": 38585952} +{"current_steps": 38360, "total_steps": 42420, "loss": 0.0402, "lr": 1.3829425012941737e-06, "epoch": 18.085808580858085, "percentage": 90.43, "elapsed_time": "2:18:58", "remaining_time": "0:14:42", "throughput": 4628.07, "total_tokens": 38590528} +{"current_steps": 38365, "total_steps": 42420, "loss": 0.0429, "lr": 1.3795708264802642e-06, "epoch": 18.08816595945309, "percentage": 90.44, "elapsed_time": "2:18:59", "remaining_time": "0:14:41", "throughput": 4628.05, "total_tokens": 38594656} +{"current_steps": 38370, "total_steps": 42420, "loss": 0.0563, "lr": 1.3762031502039613e-06, "epoch": 18.09052333804809, "percentage": 90.45, "elapsed_time": "2:19:00", "remaining_time": "0:14:40", "throughput": 4628.08, "total_tokens": 38599712} +{"current_steps": 38375, "total_steps": 42420, "loss": 0.2606, "lr": 1.3728394730353562e-06, "epoch": 18.092880716643094, "percentage": 90.46, "elapsed_time": "2:19:01", "remaining_time": "0:14:39", "throughput": 4628.11, "total_tokens": 38604832} +{"current_steps": 38380, "total_steps": 42420, "loss": 0.0848, "lr": 1.3694797955438655e-06, "epoch": 18.095238095238095, "percentage": 90.48, "elapsed_time": "2:19:02", "remaining_time": "0:14:38", "throughput": 4628.11, "total_tokens": 38609088} +{"current_steps": 38385, "total_steps": 42420, "loss": 0.1127, "lr": 1.3661241182982176e-06, "epoch": 18.0975954738331, "percentage": 90.49, "elapsed_time": "2:19:03", "remaining_time": "0:14:37", "throughput": 4628.14, "total_tokens": 38613824} +{"current_steps": 38390, "total_steps": 42420, "loss": 0.1049, "lr": 1.3627724418664778e-06, "epoch": 18.0999528524281, "percentage": 90.5, "elapsed_time": "2:19:04", "remaining_time": "0:14:35", "throughput": 4628.16, "total_tokens": 38618752} +{"current_steps": 38395, "total_steps": 42420, "loss": 0.1734, "lr": 1.3594247668160253e-06, "epoch": 18.102310231023104, "percentage": 90.51, "elapsed_time": "2:19:05", "remaining_time": "0:14:34", "throughput": 4628.19, "total_tokens": 38624000} +{"current_steps": 38400, "total_steps": 42420, "loss": 0.0639, "lr": 1.3560810937135655e-06, "epoch": 18.104667609618105, "percentage": 90.52, "elapsed_time": "2:19:06", "remaining_time": "0:14:33", "throughput": 4628.2, "total_tokens": 38628256} +{"current_steps": 38405, "total_steps": 42420, "loss": 0.2052, "lr": 1.3527414231251256e-06, "epoch": 18.10702498821311, "percentage": 90.54, "elapsed_time": "2:19:07", "remaining_time": "0:14:32", "throughput": 4628.23, "total_tokens": 38633504} +{"current_steps": 38410, "total_steps": 42420, "loss": 0.1238, "lr": 1.3494057556160566e-06, "epoch": 18.10938236680811, "percentage": 90.55, "elapsed_time": "2:19:08", "remaining_time": "0:14:31", "throughput": 4628.21, "total_tokens": 38637728} +{"current_steps": 38415, "total_steps": 42420, "loss": 0.0541, "lr": 1.346074091751029e-06, "epoch": 18.111739745403113, "percentage": 90.56, "elapsed_time": "2:19:09", "remaining_time": "0:14:30", "throughput": 4628.21, "total_tokens": 38642624} +{"current_steps": 38420, "total_steps": 42420, "loss": 0.0856, "lr": 1.3427464320940386e-06, "epoch": 18.114097123998114, "percentage": 90.57, "elapsed_time": "2:19:10", "remaining_time": "0:14:29", "throughput": 4628.2, "total_tokens": 38647040} +{"current_steps": 38425, "total_steps": 42420, "loss": 0.0855, "lr": 1.3394227772083961e-06, "epoch": 18.116454502593115, "percentage": 90.58, "elapsed_time": "2:19:11", "remaining_time": "0:14:28", "throughput": 4628.17, "total_tokens": 38650784} +{"current_steps": 38430, "total_steps": 42420, "loss": 0.1547, "lr": 1.336103127656746e-06, "epoch": 18.11881188118812, "percentage": 90.59, "elapsed_time": "2:19:12", "remaining_time": "0:14:27", "throughput": 4628.19, "total_tokens": 38656224} +{"current_steps": 38435, "total_steps": 42420, "loss": 0.0317, "lr": 1.332787484001044e-06, "epoch": 18.12116925978312, "percentage": 90.61, "elapsed_time": "2:19:13", "remaining_time": "0:14:26", "throughput": 4628.23, "total_tokens": 38661952} +{"current_steps": 38440, "total_steps": 42420, "loss": 0.148, "lr": 1.3294758468025802e-06, "epoch": 18.123526638378124, "percentage": 90.62, "elapsed_time": "2:19:14", "remaining_time": "0:14:25", "throughput": 4628.25, "total_tokens": 38666816} +{"current_steps": 38445, "total_steps": 42420, "loss": 0.0567, "lr": 1.326168216621948e-06, "epoch": 18.125884016973124, "percentage": 90.63, "elapsed_time": "2:19:15", "remaining_time": "0:14:23", "throughput": 4628.25, "total_tokens": 38670720} +{"current_steps": 38450, "total_steps": 42420, "loss": 0.0456, "lr": 1.322864594019077e-06, "epoch": 18.12824139556813, "percentage": 90.64, "elapsed_time": "2:19:16", "remaining_time": "0:14:22", "throughput": 4628.28, "total_tokens": 38675168} +{"current_steps": 38455, "total_steps": 42420, "loss": 0.1743, "lr": 1.3195649795532173e-06, "epoch": 18.13059877416313, "percentage": 90.65, "elapsed_time": "2:19:17", "remaining_time": "0:14:21", "throughput": 4628.3, "total_tokens": 38680512} +{"current_steps": 38460, "total_steps": 42420, "loss": 0.0605, "lr": 1.3162693737829356e-06, "epoch": 18.132956152758133, "percentage": 90.66, "elapsed_time": "2:19:18", "remaining_time": "0:14:20", "throughput": 4628.32, "total_tokens": 38685248} +{"current_steps": 38465, "total_steps": 42420, "loss": 0.1778, "lr": 1.312977777266125e-06, "epoch": 18.135313531353134, "percentage": 90.68, "elapsed_time": "2:19:19", "remaining_time": "0:14:19", "throughput": 4628.36, "total_tokens": 38691136} +{"current_steps": 38470, "total_steps": 42420, "loss": 0.0664, "lr": 1.3096901905599895e-06, "epoch": 18.137670909948138, "percentage": 90.69, "elapsed_time": "2:19:20", "remaining_time": "0:14:18", "throughput": 4628.38, "total_tokens": 38695872} +{"current_steps": 38475, "total_steps": 42420, "loss": 0.0437, "lr": 1.3064066142210673e-06, "epoch": 18.14002828854314, "percentage": 90.7, "elapsed_time": "2:19:21", "remaining_time": "0:14:17", "throughput": 4628.4, "total_tokens": 38700672} +{"current_steps": 38480, "total_steps": 42420, "loss": 0.0862, "lr": 1.303127048805211e-06, "epoch": 18.142385667138143, "percentage": 90.71, "elapsed_time": "2:19:22", "remaining_time": "0:14:16", "throughput": 4628.45, "total_tokens": 38706368} +{"current_steps": 38485, "total_steps": 42420, "loss": 0.0878, "lr": 1.2998514948675932e-06, "epoch": 18.144743045733144, "percentage": 90.72, "elapsed_time": "2:19:23", "remaining_time": "0:14:15", "throughput": 4628.48, "total_tokens": 38711360} +{"current_steps": 38490, "total_steps": 42420, "loss": 0.1077, "lr": 1.296579952962715e-06, "epoch": 18.147100424328148, "percentage": 90.74, "elapsed_time": "2:19:24", "remaining_time": "0:14:14", "throughput": 4628.49, "total_tokens": 38715840} +{"current_steps": 38495, "total_steps": 42420, "loss": 0.2678, "lr": 1.2933124236443884e-06, "epoch": 18.14945780292315, "percentage": 90.75, "elapsed_time": "2:19:25", "remaining_time": "0:14:12", "throughput": 4628.52, "total_tokens": 38720608} +{"current_steps": 38500, "total_steps": 42420, "loss": 0.117, "lr": 1.2900489074657519e-06, "epoch": 18.151815181518153, "percentage": 90.76, "elapsed_time": "2:19:26", "remaining_time": "0:14:11", "throughput": 4628.55, "total_tokens": 38725440} +{"current_steps": 38505, "total_steps": 42420, "loss": 0.0777, "lr": 1.2867894049792662e-06, "epoch": 18.154172560113153, "percentage": 90.77, "elapsed_time": "2:19:27", "remaining_time": "0:14:10", "throughput": 4628.53, "total_tokens": 38729376} +{"current_steps": 38510, "total_steps": 42420, "loss": 0.1746, "lr": 1.2835339167367038e-06, "epoch": 18.156529938708157, "percentage": 90.78, "elapsed_time": "2:19:28", "remaining_time": "0:14:09", "throughput": 4628.56, "total_tokens": 38734816} +{"current_steps": 38515, "total_steps": 42420, "loss": 0.1885, "lr": 1.2802824432891658e-06, "epoch": 18.158887317303158, "percentage": 90.79, "elapsed_time": "2:19:29", "remaining_time": "0:14:08", "throughput": 4628.61, "total_tokens": 38740800} +{"current_steps": 38520, "total_steps": 42420, "loss": 0.064, "lr": 1.2770349851870727e-06, "epoch": 18.161244695898162, "percentage": 90.81, "elapsed_time": "2:19:30", "remaining_time": "0:14:07", "throughput": 4628.62, "total_tokens": 38745024} +{"current_steps": 38525, "total_steps": 42420, "loss": 0.0423, "lr": 1.2737915429801628e-06, "epoch": 18.163602074493163, "percentage": 90.82, "elapsed_time": "2:19:31", "remaining_time": "0:14:06", "throughput": 4628.63, "total_tokens": 38749632} +{"current_steps": 38530, "total_steps": 42420, "loss": 0.1186, "lr": 1.2705521172174994e-06, "epoch": 18.165959453088167, "percentage": 90.83, "elapsed_time": "2:19:32", "remaining_time": "0:14:05", "throughput": 4628.65, "total_tokens": 38754464} +{"current_steps": 38535, "total_steps": 42420, "loss": 0.0427, "lr": 1.267316708447458e-06, "epoch": 18.168316831683168, "percentage": 90.84, "elapsed_time": "2:19:33", "remaining_time": "0:14:04", "throughput": 4628.69, "total_tokens": 38760032} +{"current_steps": 38540, "total_steps": 42420, "loss": 0.0707, "lr": 1.2640853172177392e-06, "epoch": 18.17067421027817, "percentage": 90.85, "elapsed_time": "2:19:34", "remaining_time": "0:14:03", "throughput": 4628.72, "total_tokens": 38764896} +{"current_steps": 38545, "total_steps": 42420, "loss": 0.0995, "lr": 1.2608579440753664e-06, "epoch": 18.173031588873172, "percentage": 90.87, "elapsed_time": "2:19:35", "remaining_time": "0:14:02", "throughput": 4628.75, "total_tokens": 38769760} +{"current_steps": 38550, "total_steps": 42420, "loss": 0.1035, "lr": 1.2576345895666725e-06, "epoch": 18.175388967468177, "percentage": 90.88, "elapsed_time": "2:19:36", "remaining_time": "0:14:00", "throughput": 4628.78, "total_tokens": 38774976} +{"current_steps": 38555, "total_steps": 42420, "loss": 0.0762, "lr": 1.2544152542373205e-06, "epoch": 18.177746346063177, "percentage": 90.89, "elapsed_time": "2:19:37", "remaining_time": "0:13:59", "throughput": 4628.79, "total_tokens": 38779392} +{"current_steps": 38560, "total_steps": 42420, "loss": 0.0641, "lr": 1.2511999386322914e-06, "epoch": 18.18010372465818, "percentage": 90.9, "elapsed_time": "2:19:38", "remaining_time": "0:13:58", "throughput": 4628.81, "total_tokens": 38784512} +{"current_steps": 38565, "total_steps": 42420, "loss": 0.1706, "lr": 1.2479886432958777e-06, "epoch": 18.182461103253182, "percentage": 90.91, "elapsed_time": "2:19:39", "remaining_time": "0:13:57", "throughput": 4628.81, "total_tokens": 38788736} +{"current_steps": 38570, "total_steps": 42420, "loss": 0.0556, "lr": 1.2447813687717024e-06, "epoch": 18.184818481848186, "percentage": 90.92, "elapsed_time": "2:19:40", "remaining_time": "0:13:56", "throughput": 4628.82, "total_tokens": 38793376} +{"current_steps": 38575, "total_steps": 42420, "loss": 0.0929, "lr": 1.2415781156026985e-06, "epoch": 18.187175860443187, "percentage": 90.94, "elapsed_time": "2:19:41", "remaining_time": "0:13:55", "throughput": 4628.85, "total_tokens": 38798656} +{"current_steps": 38580, "total_steps": 42420, "loss": 0.0944, "lr": 1.238378884331126e-06, "epoch": 18.18953323903819, "percentage": 90.95, "elapsed_time": "2:19:42", "remaining_time": "0:13:54", "throughput": 4628.87, "total_tokens": 38803584} +{"current_steps": 38585, "total_steps": 42420, "loss": 0.1231, "lr": 1.2351836754985636e-06, "epoch": 18.19189061763319, "percentage": 90.96, "elapsed_time": "2:19:43", "remaining_time": "0:13:53", "throughput": 4628.9, "total_tokens": 38808352} +{"current_steps": 38590, "total_steps": 42420, "loss": 0.0884, "lr": 1.2319924896458973e-06, "epoch": 18.194247996228196, "percentage": 90.97, "elapsed_time": "2:19:45", "remaining_time": "0:13:52", "throughput": 4628.94, "total_tokens": 38813696} +{"current_steps": 38595, "total_steps": 42420, "loss": 0.1087, "lr": 1.2288053273133426e-06, "epoch": 18.196605374823196, "percentage": 90.98, "elapsed_time": "2:19:46", "remaining_time": "0:13:51", "throughput": 4629.0, "total_tokens": 38819776} +{"current_steps": 38600, "total_steps": 42420, "loss": 0.0214, "lr": 1.2256221890404374e-06, "epoch": 18.1989627534182, "percentage": 90.99, "elapsed_time": "2:19:47", "remaining_time": "0:13:50", "throughput": 4629.03, "total_tokens": 38825152} +{"current_steps": 38605, "total_steps": 42420, "loss": 0.0234, "lr": 1.2224430753660282e-06, "epoch": 18.2013201320132, "percentage": 91.01, "elapsed_time": "2:19:48", "remaining_time": "0:13:48", "throughput": 4629.04, "total_tokens": 38829760} +{"current_steps": 38610, "total_steps": 42420, "loss": 0.0265, "lr": 1.2192679868282848e-06, "epoch": 18.203677510608205, "percentage": 91.02, "elapsed_time": "2:19:49", "remaining_time": "0:13:47", "throughput": 4629.07, "total_tokens": 38834528} +{"current_steps": 38615, "total_steps": 42420, "loss": 0.1471, "lr": 1.216096923964699e-06, "epoch": 18.206034889203206, "percentage": 91.03, "elapsed_time": "2:19:50", "remaining_time": "0:13:46", "throughput": 4629.09, "total_tokens": 38839296} +{"current_steps": 38620, "total_steps": 42420, "loss": 0.1467, "lr": 1.2129298873120753e-06, "epoch": 18.208392267798207, "percentage": 91.04, "elapsed_time": "2:19:51", "remaining_time": "0:13:45", "throughput": 4629.12, "total_tokens": 38844192} +{"current_steps": 38625, "total_steps": 42420, "loss": 0.2066, "lr": 1.2097668774065458e-06, "epoch": 18.21074964639321, "percentage": 91.05, "elapsed_time": "2:19:52", "remaining_time": "0:13:44", "throughput": 4629.14, "total_tokens": 38849440} +{"current_steps": 38630, "total_steps": 42420, "loss": 0.0273, "lr": 1.206607894783543e-06, "epoch": 18.21310702498821, "percentage": 91.07, "elapsed_time": "2:19:53", "remaining_time": "0:13:43", "throughput": 4629.17, "total_tokens": 38854432} +{"current_steps": 38635, "total_steps": 42420, "loss": 0.1022, "lr": 1.2034529399778365e-06, "epoch": 18.215464403583216, "percentage": 91.08, "elapsed_time": "2:19:54", "remaining_time": "0:13:42", "throughput": 4629.19, "total_tokens": 38859744} +{"current_steps": 38640, "total_steps": 42420, "loss": 0.0428, "lr": 1.2003020135235021e-06, "epoch": 18.217821782178216, "percentage": 91.09, "elapsed_time": "2:19:55", "remaining_time": "0:13:41", "throughput": 4629.22, "total_tokens": 38864672} +{"current_steps": 38645, "total_steps": 42420, "loss": 0.0976, "lr": 1.1971551159539434e-06, "epoch": 18.22017916077322, "percentage": 91.1, "elapsed_time": "2:19:56", "remaining_time": "0:13:40", "throughput": 4629.25, "total_tokens": 38869760} +{"current_steps": 38650, "total_steps": 42420, "loss": 0.1109, "lr": 1.1940122478018734e-06, "epoch": 18.22253653936822, "percentage": 91.11, "elapsed_time": "2:19:57", "remaining_time": "0:13:39", "throughput": 4629.27, "total_tokens": 38874592} +{"current_steps": 38655, "total_steps": 42420, "loss": 0.1565, "lr": 1.1908734095993274e-06, "epoch": 18.224893917963225, "percentage": 91.12, "elapsed_time": "2:19:58", "remaining_time": "0:13:38", "throughput": 4629.28, "total_tokens": 38879008} +{"current_steps": 38660, "total_steps": 42420, "loss": 0.0397, "lr": 1.187738601877658e-06, "epoch": 18.227251296558226, "percentage": 91.14, "elapsed_time": "2:19:59", "remaining_time": "0:13:36", "throughput": 4629.31, "total_tokens": 38883680} +{"current_steps": 38665, "total_steps": 42420, "loss": 0.1841, "lr": 1.1846078251675386e-06, "epoch": 18.22960867515323, "percentage": 91.15, "elapsed_time": "2:20:00", "remaining_time": "0:13:35", "throughput": 4629.34, "total_tokens": 38889280} +{"current_steps": 38670, "total_steps": 42420, "loss": 0.0801, "lr": 1.1814810799989474e-06, "epoch": 18.23196605374823, "percentage": 91.16, "elapsed_time": "2:20:01", "remaining_time": "0:13:34", "throughput": 4629.38, "total_tokens": 38894464} +{"current_steps": 38675, "total_steps": 42420, "loss": 0.2402, "lr": 1.1783583669011943e-06, "epoch": 18.234323432343235, "percentage": 91.17, "elapsed_time": "2:20:02", "remaining_time": "0:13:33", "throughput": 4629.41, "total_tokens": 38899520} +{"current_steps": 38680, "total_steps": 42420, "loss": 0.1279, "lr": 1.1752396864029013e-06, "epoch": 18.236680810938235, "percentage": 91.18, "elapsed_time": "2:20:03", "remaining_time": "0:13:32", "throughput": 4629.45, "total_tokens": 38904896} +{"current_steps": 38685, "total_steps": 42420, "loss": 0.1104, "lr": 1.172125039032007e-06, "epoch": 18.23903818953324, "percentage": 91.2, "elapsed_time": "2:20:04", "remaining_time": "0:13:31", "throughput": 4629.48, "total_tokens": 38910048} +{"current_steps": 38690, "total_steps": 42420, "loss": 0.0703, "lr": 1.1690144253157704e-06, "epoch": 18.24139556812824, "percentage": 91.21, "elapsed_time": "2:20:05", "remaining_time": "0:13:30", "throughput": 4629.46, "total_tokens": 38913952} +{"current_steps": 38695, "total_steps": 42420, "loss": 0.1583, "lr": 1.1659078457807644e-06, "epoch": 18.243752946723244, "percentage": 91.22, "elapsed_time": "2:20:07", "remaining_time": "0:13:29", "throughput": 4629.53, "total_tokens": 38921216} +{"current_steps": 38700, "total_steps": 42420, "loss": 0.0849, "lr": 1.1628053009528766e-06, "epoch": 18.246110325318245, "percentage": 91.23, "elapsed_time": "2:20:08", "remaining_time": "0:13:28", "throughput": 4629.55, "total_tokens": 38925696} +{"current_steps": 38705, "total_steps": 42420, "loss": 0.1707, "lr": 1.1597067913573234e-06, "epoch": 18.24846770391325, "percentage": 91.24, "elapsed_time": "2:20:09", "remaining_time": "0:13:27", "throughput": 4629.6, "total_tokens": 38932256} +{"current_steps": 38710, "total_steps": 42420, "loss": 0.1796, "lr": 1.1566123175186238e-06, "epoch": 18.25082508250825, "percentage": 91.25, "elapsed_time": "2:20:10", "remaining_time": "0:13:26", "throughput": 4629.62, "total_tokens": 38937248} +{"current_steps": 38715, "total_steps": 42420, "loss": 0.0563, "lr": 1.153521879960623e-06, "epoch": 18.253182461103254, "percentage": 91.27, "elapsed_time": "2:20:11", "remaining_time": "0:13:24", "throughput": 4629.66, "total_tokens": 38942720} +{"current_steps": 38720, "total_steps": 42420, "loss": 0.1747, "lr": 1.150435479206477e-06, "epoch": 18.255539839698255, "percentage": 91.28, "elapsed_time": "2:20:13", "remaining_time": "0:13:23", "throughput": 4629.73, "total_tokens": 38952256} +{"current_steps": 38725, "total_steps": 42420, "loss": 0.1345, "lr": 1.14735311577866e-06, "epoch": 18.25789721829326, "percentage": 91.29, "elapsed_time": "2:20:14", "remaining_time": "0:13:22", "throughput": 4629.75, "total_tokens": 38956640} +{"current_steps": 38730, "total_steps": 42420, "loss": 0.1631, "lr": 1.1442747901989653e-06, "epoch": 18.26025459688826, "percentage": 91.3, "elapsed_time": "2:20:15", "remaining_time": "0:13:21", "throughput": 4629.77, "total_tokens": 38961088} +{"current_steps": 38735, "total_steps": 42420, "loss": 0.095, "lr": 1.1412005029885015e-06, "epoch": 18.262611975483264, "percentage": 91.31, "elapsed_time": "2:20:16", "remaining_time": "0:13:20", "throughput": 4629.78, "total_tokens": 38965952} +{"current_steps": 38740, "total_steps": 42420, "loss": 0.1468, "lr": 1.1381302546676936e-06, "epoch": 18.264969354078264, "percentage": 91.32, "elapsed_time": "2:20:17", "remaining_time": "0:13:19", "throughput": 4629.79, "total_tokens": 38970528} +{"current_steps": 38745, "total_steps": 42420, "loss": 0.2125, "lr": 1.1350640457562844e-06, "epoch": 18.26732673267327, "percentage": 91.34, "elapsed_time": "2:20:18", "remaining_time": "0:13:18", "throughput": 4629.8, "total_tokens": 38974784} +{"current_steps": 38750, "total_steps": 42420, "loss": 0.0359, "lr": 1.1320018767733254e-06, "epoch": 18.26968411126827, "percentage": 91.35, "elapsed_time": "2:20:19", "remaining_time": "0:13:17", "throughput": 4629.77, "total_tokens": 38978912} +{"current_steps": 38755, "total_steps": 42420, "loss": 0.0374, "lr": 1.1289437482371989e-06, "epoch": 18.272041489863273, "percentage": 91.36, "elapsed_time": "2:20:20", "remaining_time": "0:13:16", "throughput": 4629.79, "total_tokens": 38983808} +{"current_steps": 38760, "total_steps": 42420, "loss": 0.1974, "lr": 1.1258896606655827e-06, "epoch": 18.274398868458274, "percentage": 91.37, "elapsed_time": "2:20:21", "remaining_time": "0:13:15", "throughput": 4629.79, "total_tokens": 38988544} +{"current_steps": 38765, "total_steps": 42420, "loss": 0.1034, "lr": 1.1228396145754882e-06, "epoch": 18.276756247053278, "percentage": 91.38, "elapsed_time": "2:20:22", "remaining_time": "0:13:14", "throughput": 4629.81, "total_tokens": 38993696} +{"current_steps": 38770, "total_steps": 42420, "loss": 0.1283, "lr": 1.1197936104832384e-06, "epoch": 18.27911362564828, "percentage": 91.4, "elapsed_time": "2:20:23", "remaining_time": "0:13:13", "throughput": 4629.8, "total_tokens": 38998240} +{"current_steps": 38775, "total_steps": 42420, "loss": 0.1337, "lr": 1.1167516489044656e-06, "epoch": 18.281471004243283, "percentage": 91.41, "elapsed_time": "2:20:24", "remaining_time": "0:13:11", "throughput": 4629.84, "total_tokens": 39004416} +{"current_steps": 38780, "total_steps": 42420, "loss": 0.0329, "lr": 1.113713730354124e-06, "epoch": 18.283828382838283, "percentage": 91.42, "elapsed_time": "2:20:25", "remaining_time": "0:13:10", "throughput": 4629.86, "total_tokens": 39009440} +{"current_steps": 38785, "total_steps": 42420, "loss": 0.1089, "lr": 1.1106798553464804e-06, "epoch": 18.286185761433288, "percentage": 91.43, "elapsed_time": "2:20:26", "remaining_time": "0:13:09", "throughput": 4629.86, "total_tokens": 39014016} +{"current_steps": 38790, "total_steps": 42420, "loss": 0.0405, "lr": 1.107650024395121e-06, "epoch": 18.28854314002829, "percentage": 91.44, "elapsed_time": "2:20:27", "remaining_time": "0:13:08", "throughput": 4629.89, "total_tokens": 39018944} +{"current_steps": 38795, "total_steps": 42420, "loss": 0.1871, "lr": 1.1046242380129495e-06, "epoch": 18.290900518623292, "percentage": 91.45, "elapsed_time": "2:20:28", "remaining_time": "0:13:07", "throughput": 4629.92, "total_tokens": 39024480} +{"current_steps": 38800, "total_steps": 42420, "loss": 0.3587, "lr": 1.1016024967121674e-06, "epoch": 18.293257897218293, "percentage": 91.47, "elapsed_time": "2:20:29", "remaining_time": "0:13:06", "throughput": 4629.95, "total_tokens": 39029536} +{"current_steps": 38805, "total_steps": 42420, "loss": 0.0605, "lr": 1.0985848010043126e-06, "epoch": 18.295615275813297, "percentage": 91.48, "elapsed_time": "2:20:30", "remaining_time": "0:13:05", "throughput": 4629.98, "total_tokens": 39034400} +{"current_steps": 38810, "total_steps": 42420, "loss": 0.2949, "lr": 1.0955711514002265e-06, "epoch": 18.297972654408298, "percentage": 91.49, "elapsed_time": "2:20:31", "remaining_time": "0:13:04", "throughput": 4630.01, "total_tokens": 39039616} +{"current_steps": 38815, "total_steps": 42420, "loss": 0.1915, "lr": 1.0925615484100704e-06, "epoch": 18.300330033003302, "percentage": 91.5, "elapsed_time": "2:20:32", "remaining_time": "0:13:03", "throughput": 4630.03, "total_tokens": 39044384} +{"current_steps": 38820, "total_steps": 42420, "loss": 0.0663, "lr": 1.0895559925433201e-06, "epoch": 18.302687411598303, "percentage": 91.51, "elapsed_time": "2:20:33", "remaining_time": "0:13:02", "throughput": 4630.06, "total_tokens": 39049824} +{"current_steps": 38825, "total_steps": 42420, "loss": 0.0252, "lr": 1.0865544843087634e-06, "epoch": 18.305044790193303, "percentage": 91.53, "elapsed_time": "2:20:34", "remaining_time": "0:13:01", "throughput": 4630.06, "total_tokens": 39054240} +{"current_steps": 38830, "total_steps": 42420, "loss": 0.1493, "lr": 1.0835570242145071e-06, "epoch": 18.307402168788308, "percentage": 91.54, "elapsed_time": "2:20:35", "remaining_time": "0:12:59", "throughput": 4630.09, "total_tokens": 39059168} +{"current_steps": 38835, "total_steps": 42420, "loss": 0.0618, "lr": 1.080563612767971e-06, "epoch": 18.309759547383308, "percentage": 91.55, "elapsed_time": "2:20:36", "remaining_time": "0:12:58", "throughput": 4630.09, "total_tokens": 39063296} +{"current_steps": 38840, "total_steps": 42420, "loss": 0.0932, "lr": 1.0775742504758795e-06, "epoch": 18.312116925978312, "percentage": 91.56, "elapsed_time": "2:20:37", "remaining_time": "0:12:57", "throughput": 4630.11, "total_tokens": 39067968} +{"current_steps": 38845, "total_steps": 42420, "loss": 0.0968, "lr": 1.0745889378442924e-06, "epoch": 18.314474304573313, "percentage": 91.57, "elapsed_time": "2:20:38", "remaining_time": "0:12:56", "throughput": 4630.11, "total_tokens": 39072480} +{"current_steps": 38850, "total_steps": 42420, "loss": 0.0499, "lr": 1.0716076753785664e-06, "epoch": 18.316831683168317, "percentage": 91.58, "elapsed_time": "2:20:39", "remaining_time": "0:12:55", "throughput": 4630.15, "total_tokens": 39077696} +{"current_steps": 38855, "total_steps": 42420, "loss": 0.0247, "lr": 1.0686304635833782e-06, "epoch": 18.319189061763318, "percentage": 91.6, "elapsed_time": "2:20:40", "remaining_time": "0:12:54", "throughput": 4630.16, "total_tokens": 39082048} +{"current_steps": 38860, "total_steps": 42420, "loss": 0.1426, "lr": 1.0656573029627221e-06, "epoch": 18.321546440358322, "percentage": 91.61, "elapsed_time": "2:20:41", "remaining_time": "0:12:53", "throughput": 4630.19, "total_tokens": 39087392} +{"current_steps": 38865, "total_steps": 42420, "loss": 0.093, "lr": 1.0626881940199035e-06, "epoch": 18.323903818953323, "percentage": 91.62, "elapsed_time": "2:20:42", "remaining_time": "0:12:52", "throughput": 4630.21, "total_tokens": 39091552} +{"current_steps": 38870, "total_steps": 42420, "loss": 0.12, "lr": 1.0597231372575401e-06, "epoch": 18.326261197548327, "percentage": 91.63, "elapsed_time": "2:20:43", "remaining_time": "0:12:51", "throughput": 4630.24, "total_tokens": 39096448} +{"current_steps": 38875, "total_steps": 42420, "loss": 0.1747, "lr": 1.0567621331775717e-06, "epoch": 18.328618576143327, "percentage": 91.64, "elapsed_time": "2:20:44", "remaining_time": "0:12:50", "throughput": 4630.23, "total_tokens": 39100480} +{"current_steps": 38880, "total_steps": 42420, "loss": 0.1812, "lr": 1.0538051822812367e-06, "epoch": 18.33097595473833, "percentage": 91.65, "elapsed_time": "2:20:45", "remaining_time": "0:12:48", "throughput": 4630.26, "total_tokens": 39105664} +{"current_steps": 38885, "total_steps": 42420, "loss": 0.0949, "lr": 1.0508522850691012e-06, "epoch": 18.333333333333332, "percentage": 91.67, "elapsed_time": "2:20:46", "remaining_time": "0:12:47", "throughput": 4630.3, "total_tokens": 39111104} +{"current_steps": 38890, "total_steps": 42420, "loss": 0.0876, "lr": 1.0479034420410427e-06, "epoch": 18.335690711928336, "percentage": 91.68, "elapsed_time": "2:20:47", "remaining_time": "0:12:46", "throughput": 4630.3, "total_tokens": 39115680} +{"current_steps": 38895, "total_steps": 42420, "loss": 0.1097, "lr": 1.0449586536962458e-06, "epoch": 18.338048090523337, "percentage": 91.69, "elapsed_time": "2:20:48", "remaining_time": "0:12:45", "throughput": 4630.33, "total_tokens": 39120992} +{"current_steps": 38900, "total_steps": 42420, "loss": 0.1196, "lr": 1.0420179205332142e-06, "epoch": 18.34040546911834, "percentage": 91.7, "elapsed_time": "2:20:50", "remaining_time": "0:12:44", "throughput": 4630.37, "total_tokens": 39127104} +{"current_steps": 38905, "total_steps": 42420, "loss": 0.0424, "lr": 1.0390812430497665e-06, "epoch": 18.34276284771334, "percentage": 91.71, "elapsed_time": "2:20:51", "remaining_time": "0:12:43", "throughput": 4630.39, "total_tokens": 39132256} +{"current_steps": 38910, "total_steps": 42420, "loss": 0.0496, "lr": 1.0361486217430328e-06, "epoch": 18.345120226308346, "percentage": 91.73, "elapsed_time": "2:20:52", "remaining_time": "0:12:42", "throughput": 4630.4, "total_tokens": 39137504} +{"current_steps": 38915, "total_steps": 42420, "loss": 0.1184, "lr": 1.0332200571094518e-06, "epoch": 18.347477604903347, "percentage": 91.74, "elapsed_time": "2:20:53", "remaining_time": "0:12:41", "throughput": 4630.41, "total_tokens": 39142368} +{"current_steps": 38920, "total_steps": 42420, "loss": 0.1596, "lr": 1.030295549644783e-06, "epoch": 18.34983498349835, "percentage": 91.75, "elapsed_time": "2:20:54", "remaining_time": "0:12:40", "throughput": 4630.44, "total_tokens": 39147072} +{"current_steps": 38925, "total_steps": 42420, "loss": 0.2228, "lr": 1.0273750998440996e-06, "epoch": 18.35219236209335, "percentage": 91.76, "elapsed_time": "2:20:55", "remaining_time": "0:12:39", "throughput": 4630.46, "total_tokens": 39151680} +{"current_steps": 38930, "total_steps": 42420, "loss": 0.0987, "lr": 1.0244587082017753e-06, "epoch": 18.354549740688356, "percentage": 91.77, "elapsed_time": "2:20:56", "remaining_time": "0:12:38", "throughput": 4630.48, "total_tokens": 39156032} +{"current_steps": 38935, "total_steps": 42420, "loss": 0.1703, "lr": 1.0215463752115101e-06, "epoch": 18.356907119283356, "percentage": 91.78, "elapsed_time": "2:20:57", "remaining_time": "0:12:37", "throughput": 4630.52, "total_tokens": 39162912} +{"current_steps": 38940, "total_steps": 42420, "loss": 0.1928, "lr": 1.018638101366312e-06, "epoch": 18.35926449787836, "percentage": 91.8, "elapsed_time": "2:20:58", "remaining_time": "0:12:35", "throughput": 4630.55, "total_tokens": 39168512} +{"current_steps": 38945, "total_steps": 42420, "loss": 0.0606, "lr": 1.0157338871585043e-06, "epoch": 18.36162187647336, "percentage": 91.81, "elapsed_time": "2:20:59", "remaining_time": "0:12:34", "throughput": 4630.58, "total_tokens": 39173696} +{"current_steps": 38950, "total_steps": 42420, "loss": 0.0343, "lr": 1.0128337330797182e-06, "epoch": 18.363979255068365, "percentage": 91.82, "elapsed_time": "2:21:00", "remaining_time": "0:12:33", "throughput": 4630.58, "total_tokens": 39177664} +{"current_steps": 38955, "total_steps": 42420, "loss": 0.1743, "lr": 1.0099376396209032e-06, "epoch": 18.366336633663366, "percentage": 91.83, "elapsed_time": "2:21:01", "remaining_time": "0:12:32", "throughput": 4630.6, "total_tokens": 39182400} +{"current_steps": 38960, "total_steps": 42420, "loss": 0.122, "lr": 1.0070456072723194e-06, "epoch": 18.36869401225837, "percentage": 91.84, "elapsed_time": "2:21:02", "remaining_time": "0:12:31", "throughput": 4630.6, "total_tokens": 39186976} +{"current_steps": 38965, "total_steps": 42420, "loss": 0.0136, "lr": 1.0041576365235395e-06, "epoch": 18.37105139085337, "percentage": 91.86, "elapsed_time": "2:21:03", "remaining_time": "0:12:30", "throughput": 4630.59, "total_tokens": 39190976} +{"current_steps": 38970, "total_steps": 42420, "loss": 0.0671, "lr": 1.0012737278634416e-06, "epoch": 18.373408769448375, "percentage": 91.87, "elapsed_time": "2:21:04", "remaining_time": "0:12:29", "throughput": 4630.66, "total_tokens": 39197504} +{"current_steps": 38975, "total_steps": 42420, "loss": 0.0333, "lr": 9.983938817802268e-07, "epoch": 18.375766148043375, "percentage": 91.88, "elapsed_time": "2:21:05", "remaining_time": "0:12:28", "throughput": 4630.69, "total_tokens": 39203104} +{"current_steps": 38980, "total_steps": 42420, "loss": 0.0674, "lr": 9.955180987614026e-07, "epoch": 18.37812352663838, "percentage": 91.89, "elapsed_time": "2:21:07", "remaining_time": "0:12:27", "throughput": 4630.75, "total_tokens": 39209472} +{"current_steps": 38985, "total_steps": 42420, "loss": 0.0823, "lr": 9.926463792937907e-07, "epoch": 18.38048090523338, "percentage": 91.9, "elapsed_time": "2:21:08", "remaining_time": "0:12:26", "throughput": 4630.77, "total_tokens": 39214464} +{"current_steps": 38990, "total_steps": 42420, "loss": 0.0696, "lr": 9.89778723863527e-07, "epoch": 18.382838283828384, "percentage": 91.91, "elapsed_time": "2:21:09", "remaining_time": "0:12:25", "throughput": 4630.79, "total_tokens": 39219200} +{"current_steps": 38995, "total_steps": 42420, "loss": 0.168, "lr": 9.869151329560565e-07, "epoch": 18.385195662423385, "percentage": 91.93, "elapsed_time": "2:21:10", "remaining_time": "0:12:23", "throughput": 4630.77, "total_tokens": 39222912} +{"current_steps": 39000, "total_steps": 42420, "loss": 0.0822, "lr": 9.840556070561335e-07, "epoch": 18.38755304101839, "percentage": 91.94, "elapsed_time": "2:21:11", "remaining_time": "0:12:22", "throughput": 4630.8, "total_tokens": 39228064} +{"current_steps": 39005, "total_steps": 42420, "loss": 0.1048, "lr": 9.812001466478315e-07, "epoch": 18.38991041961339, "percentage": 91.95, "elapsed_time": "2:21:12", "remaining_time": "0:12:21", "throughput": 4630.85, "total_tokens": 39233792} +{"current_steps": 39010, "total_steps": 42420, "loss": 0.0651, "lr": 9.783487522145251e-07, "epoch": 18.392267798208394, "percentage": 91.96, "elapsed_time": "2:21:13", "remaining_time": "0:12:20", "throughput": 4630.83, "total_tokens": 39237568} +{"current_steps": 39015, "total_steps": 42420, "loss": 0.2261, "lr": 9.755014242389115e-07, "epoch": 18.394625176803395, "percentage": 91.97, "elapsed_time": "2:21:14", "remaining_time": "0:12:19", "throughput": 4630.83, "total_tokens": 39241952} +{"current_steps": 39020, "total_steps": 42420, "loss": 0.2001, "lr": 9.726581632029912e-07, "epoch": 18.396982555398395, "percentage": 91.98, "elapsed_time": "2:21:15", "remaining_time": "0:12:18", "throughput": 4630.8, "total_tokens": 39246240} +{"current_steps": 39025, "total_steps": 42420, "loss": 0.0363, "lr": 9.69818969588085e-07, "epoch": 18.3993399339934, "percentage": 92.0, "elapsed_time": "2:21:16", "remaining_time": "0:12:17", "throughput": 4630.81, "total_tokens": 39250848} +{"current_steps": 39030, "total_steps": 42420, "loss": 0.184, "lr": 9.669838438748192e-07, "epoch": 18.4016973125884, "percentage": 92.01, "elapsed_time": "2:21:17", "remaining_time": "0:12:16", "throughput": 4630.85, "total_tokens": 39256832} +{"current_steps": 39035, "total_steps": 42420, "loss": 0.0449, "lr": 9.641527865431294e-07, "epoch": 18.404054691183404, "percentage": 92.02, "elapsed_time": "2:21:18", "remaining_time": "0:12:15", "throughput": 4630.88, "total_tokens": 39262304} +{"current_steps": 39040, "total_steps": 42420, "loss": 0.1001, "lr": 9.613257980722684e-07, "epoch": 18.406412069778405, "percentage": 92.03, "elapsed_time": "2:21:19", "remaining_time": "0:12:14", "throughput": 4630.91, "total_tokens": 39267552} +{"current_steps": 39045, "total_steps": 42420, "loss": 0.1903, "lr": 9.585028789408006e-07, "epoch": 18.40876944837341, "percentage": 92.04, "elapsed_time": "2:21:20", "remaining_time": "0:12:13", "throughput": 4630.94, "total_tokens": 39272832} +{"current_steps": 39050, "total_steps": 42420, "loss": 0.1533, "lr": 9.556840296265884e-07, "epoch": 18.41112682696841, "percentage": 92.06, "elapsed_time": "2:21:21", "remaining_time": "0:12:11", "throughput": 4630.97, "total_tokens": 39277632} +{"current_steps": 39055, "total_steps": 42420, "loss": 0.2269, "lr": 9.528692506068221e-07, "epoch": 18.413484205563414, "percentage": 92.07, "elapsed_time": "2:21:22", "remaining_time": "0:12:10", "throughput": 4631.01, "total_tokens": 39283488} +{"current_steps": 39060, "total_steps": 42420, "loss": 0.1539, "lr": 9.500585423579955e-07, "epoch": 18.415841584158414, "percentage": 92.08, "elapsed_time": "2:21:23", "remaining_time": "0:12:09", "throughput": 4631.03, "total_tokens": 39288288} +{"current_steps": 39065, "total_steps": 42420, "loss": 0.0909, "lr": 9.472519053559114e-07, "epoch": 18.41819896275342, "percentage": 92.09, "elapsed_time": "2:21:24", "remaining_time": "0:12:08", "throughput": 4631.07, "total_tokens": 39294432} +{"current_steps": 39070, "total_steps": 42420, "loss": 0.0454, "lr": 9.444493400756871e-07, "epoch": 18.42055634134842, "percentage": 92.1, "elapsed_time": "2:21:26", "remaining_time": "0:12:07", "throughput": 4631.1, "total_tokens": 39299840} +{"current_steps": 39075, "total_steps": 42420, "loss": 0.0338, "lr": 9.416508469917512e-07, "epoch": 18.422913719943423, "percentage": 92.11, "elapsed_time": "2:21:27", "remaining_time": "0:12:06", "throughput": 4631.08, "total_tokens": 39304480} +{"current_steps": 39080, "total_steps": 42420, "loss": 0.0932, "lr": 9.38856426577836e-07, "epoch": 18.425271098538424, "percentage": 92.13, "elapsed_time": "2:21:27", "remaining_time": "0:12:05", "throughput": 4631.08, "total_tokens": 39308512} +{"current_steps": 39085, "total_steps": 42420, "loss": 0.111, "lr": 9.360660793069992e-07, "epoch": 18.427628477133428, "percentage": 92.14, "elapsed_time": "2:21:28", "remaining_time": "0:12:04", "throughput": 4631.1, "total_tokens": 39313088} +{"current_steps": 39090, "total_steps": 42420, "loss": 0.1805, "lr": 9.332798056515879e-07, "epoch": 18.42998585572843, "percentage": 92.15, "elapsed_time": "2:21:29", "remaining_time": "0:12:03", "throughput": 4631.13, "total_tokens": 39317888} +{"current_steps": 39095, "total_steps": 42420, "loss": 0.4297, "lr": 9.304976060832777e-07, "epoch": 18.432343234323433, "percentage": 92.16, "elapsed_time": "2:21:30", "remaining_time": "0:12:02", "throughput": 4631.14, "total_tokens": 39322816} +{"current_steps": 39100, "total_steps": 42420, "loss": 0.0357, "lr": 9.277194810730444e-07, "epoch": 18.434700612918434, "percentage": 92.17, "elapsed_time": "2:21:32", "remaining_time": "0:12:01", "throughput": 4631.17, "total_tokens": 39328032} +{"current_steps": 39105, "total_steps": 42420, "loss": 0.0993, "lr": 9.249454310911787e-07, "epoch": 18.437057991513438, "percentage": 92.19, "elapsed_time": "2:21:32", "remaining_time": "0:11:59", "throughput": 4631.17, "total_tokens": 39332544} +{"current_steps": 39110, "total_steps": 42420, "loss": 0.0679, "lr": 9.221754566072827e-07, "epoch": 18.43941537010844, "percentage": 92.2, "elapsed_time": "2:21:34", "remaining_time": "0:11:58", "throughput": 4631.2, "total_tokens": 39337600} +{"current_steps": 39115, "total_steps": 42420, "loss": 0.0492, "lr": 9.194095580902645e-07, "epoch": 18.441772748703443, "percentage": 92.21, "elapsed_time": "2:21:35", "remaining_time": "0:11:57", "throughput": 4631.2, "total_tokens": 39342240} +{"current_steps": 39120, "total_steps": 42420, "loss": 0.1008, "lr": 9.166477360083415e-07, "epoch": 18.444130127298443, "percentage": 92.22, "elapsed_time": "2:21:36", "remaining_time": "0:11:56", "throughput": 4631.24, "total_tokens": 39347872} +{"current_steps": 39125, "total_steps": 42420, "loss": 0.0326, "lr": 9.138899908290421e-07, "epoch": 18.446487505893447, "percentage": 92.23, "elapsed_time": "2:21:37", "remaining_time": "0:11:55", "throughput": 4631.25, "total_tokens": 39352864} +{"current_steps": 39130, "total_steps": 42420, "loss": 0.1891, "lr": 9.1113632301921e-07, "epoch": 18.448844884488448, "percentage": 92.24, "elapsed_time": "2:21:38", "remaining_time": "0:11:54", "throughput": 4631.23, "total_tokens": 39357152} +{"current_steps": 39135, "total_steps": 42420, "loss": 0.0817, "lr": 9.083867330449969e-07, "epoch": 18.451202263083452, "percentage": 92.26, "elapsed_time": "2:21:39", "remaining_time": "0:11:53", "throughput": 4631.24, "total_tokens": 39362560} +{"current_steps": 39140, "total_steps": 42420, "loss": 0.2691, "lr": 9.05641221371853e-07, "epoch": 18.453559641678453, "percentage": 92.27, "elapsed_time": "2:21:40", "remaining_time": "0:11:52", "throughput": 4631.26, "total_tokens": 39367296} +{"current_steps": 39145, "total_steps": 42420, "loss": 0.113, "lr": 9.028997884645535e-07, "epoch": 18.455917020273457, "percentage": 92.28, "elapsed_time": "2:21:41", "remaining_time": "0:11:51", "throughput": 4631.28, "total_tokens": 39372576} +{"current_steps": 39150, "total_steps": 42420, "loss": 0.1895, "lr": 9.001624347871717e-07, "epoch": 18.458274398868458, "percentage": 92.29, "elapsed_time": "2:21:42", "remaining_time": "0:11:50", "throughput": 4631.31, "total_tokens": 39378464} +{"current_steps": 39155, "total_steps": 42420, "loss": 0.1012, "lr": 8.974291608030983e-07, "epoch": 18.460631777463462, "percentage": 92.3, "elapsed_time": "2:21:43", "remaining_time": "0:11:49", "throughput": 4631.32, "total_tokens": 39383392} +{"current_steps": 39160, "total_steps": 42420, "loss": 0.1174, "lr": 8.946999669750294e-07, "epoch": 18.462989156058462, "percentage": 92.31, "elapsed_time": "2:21:44", "remaining_time": "0:11:47", "throughput": 4631.32, "total_tokens": 39387776} +{"current_steps": 39165, "total_steps": 42420, "loss": 0.1123, "lr": 8.919748537649736e-07, "epoch": 18.465346534653467, "percentage": 92.33, "elapsed_time": "2:21:45", "remaining_time": "0:11:46", "throughput": 4631.34, "total_tokens": 39392480} +{"current_steps": 39170, "total_steps": 42420, "loss": 0.1172, "lr": 8.892538216342422e-07, "epoch": 18.467703913248467, "percentage": 92.34, "elapsed_time": "2:21:46", "remaining_time": "0:11:45", "throughput": 4631.36, "total_tokens": 39397344} +{"current_steps": 39175, "total_steps": 42420, "loss": 0.082, "lr": 8.86536871043464e-07, "epoch": 18.47006129184347, "percentage": 92.35, "elapsed_time": "2:21:47", "remaining_time": "0:11:44", "throughput": 4631.37, "total_tokens": 39401824} +{"current_steps": 39180, "total_steps": 42420, "loss": 0.0195, "lr": 8.838240024525685e-07, "epoch": 18.472418670438472, "percentage": 92.36, "elapsed_time": "2:21:48", "remaining_time": "0:11:43", "throughput": 4631.37, "total_tokens": 39406560} +{"current_steps": 39185, "total_steps": 42420, "loss": 0.0569, "lr": 8.811152163208019e-07, "epoch": 18.474776049033476, "percentage": 92.37, "elapsed_time": "2:21:49", "remaining_time": "0:11:42", "throughput": 4631.37, "total_tokens": 39410912} +{"current_steps": 39190, "total_steps": 42420, "loss": 0.1619, "lr": 8.784105131067116e-07, "epoch": 18.477133427628477, "percentage": 92.39, "elapsed_time": "2:21:50", "remaining_time": "0:11:41", "throughput": 4631.37, "total_tokens": 39416096} +{"current_steps": 39195, "total_steps": 42420, "loss": 0.1104, "lr": 8.75709893268159e-07, "epoch": 18.47949080622348, "percentage": 92.4, "elapsed_time": "2:21:51", "remaining_time": "0:11:40", "throughput": 4631.38, "total_tokens": 39421088} +{"current_steps": 39200, "total_steps": 42420, "loss": 0.164, "lr": 8.730133572623172e-07, "epoch": 18.48184818481848, "percentage": 92.41, "elapsed_time": "2:21:52", "remaining_time": "0:11:39", "throughput": 4631.4, "total_tokens": 39426176} +{"current_steps": 39205, "total_steps": 42420, "loss": 0.0605, "lr": 8.703209055456573e-07, "epoch": 18.484205563413486, "percentage": 92.42, "elapsed_time": "2:21:53", "remaining_time": "0:11:38", "throughput": 4631.42, "total_tokens": 39430848} +{"current_steps": 39210, "total_steps": 42420, "loss": 0.1198, "lr": 8.676325385739731e-07, "epoch": 18.486562942008486, "percentage": 92.43, "elapsed_time": "2:21:54", "remaining_time": "0:11:37", "throughput": 4631.42, "total_tokens": 39435264} +{"current_steps": 39215, "total_steps": 42420, "loss": 0.0414, "lr": 8.649482568023559e-07, "epoch": 18.48892032060349, "percentage": 92.44, "elapsed_time": "2:21:55", "remaining_time": "0:11:35", "throughput": 4631.44, "total_tokens": 39439904} +{"current_steps": 39220, "total_steps": 42420, "loss": 0.0941, "lr": 8.622680606852091e-07, "epoch": 18.49127769919849, "percentage": 92.46, "elapsed_time": "2:21:56", "remaining_time": "0:11:34", "throughput": 4631.48, "total_tokens": 39445984} +{"current_steps": 39225, "total_steps": 42420, "loss": 0.1217, "lr": 8.595919506762418e-07, "epoch": 18.493635077793492, "percentage": 92.47, "elapsed_time": "2:21:58", "remaining_time": "0:11:33", "throughput": 4631.52, "total_tokens": 39451360} +{"current_steps": 39230, "total_steps": 42420, "loss": 0.0621, "lr": 8.569199272284778e-07, "epoch": 18.495992456388496, "percentage": 92.48, "elapsed_time": "2:21:59", "remaining_time": "0:11:32", "throughput": 4631.55, "total_tokens": 39456256} +{"current_steps": 39235, "total_steps": 42420, "loss": 0.0504, "lr": 8.542519907942442e-07, "epoch": 18.498349834983497, "percentage": 92.49, "elapsed_time": "2:22:00", "remaining_time": "0:11:31", "throughput": 4631.58, "total_tokens": 39461472} +{"current_steps": 39240, "total_steps": 42420, "loss": 0.1038, "lr": 8.51588141825177e-07, "epoch": 18.5007072135785, "percentage": 92.5, "elapsed_time": "2:22:01", "remaining_time": "0:11:30", "throughput": 4631.61, "total_tokens": 39466784} +{"current_steps": 39245, "total_steps": 42420, "loss": 0.078, "lr": 8.489283807722182e-07, "epoch": 18.5030645921735, "percentage": 92.52, "elapsed_time": "2:22:02", "remaining_time": "0:11:29", "throughput": 4631.68, "total_tokens": 39474560} +{"current_steps": 39250, "total_steps": 42420, "loss": 0.049, "lr": 8.462727080856242e-07, "epoch": 18.505421970768506, "percentage": 92.53, "elapsed_time": "2:22:03", "remaining_time": "0:11:28", "throughput": 4631.7, "total_tokens": 39480032} +{"current_steps": 39255, "total_steps": 42420, "loss": 0.1415, "lr": 8.436211242149577e-07, "epoch": 18.507779349363506, "percentage": 92.54, "elapsed_time": "2:22:04", "remaining_time": "0:11:27", "throughput": 4631.73, "total_tokens": 39485088} +{"current_steps": 39260, "total_steps": 42420, "loss": 0.0926, "lr": 8.409736296090792e-07, "epoch": 18.51013672795851, "percentage": 92.55, "elapsed_time": "2:22:06", "remaining_time": "0:11:26", "throughput": 4631.77, "total_tokens": 39490784} +{"current_steps": 39265, "total_steps": 42420, "loss": 0.1875, "lr": 8.383302247161689e-07, "epoch": 18.51249410655351, "percentage": 92.56, "elapsed_time": "2:22:07", "remaining_time": "0:11:25", "throughput": 4631.79, "total_tokens": 39495488} +{"current_steps": 39270, "total_steps": 42420, "loss": 0.2041, "lr": 8.356909099837107e-07, "epoch": 18.514851485148515, "percentage": 92.57, "elapsed_time": "2:22:08", "remaining_time": "0:11:24", "throughput": 4631.82, "total_tokens": 39500512} +{"current_steps": 39275, "total_steps": 42420, "loss": 0.0673, "lr": 8.330556858584915e-07, "epoch": 18.517208863743516, "percentage": 92.59, "elapsed_time": "2:22:08", "remaining_time": "0:11:22", "throughput": 4631.77, "total_tokens": 39503840} +{"current_steps": 39280, "total_steps": 42420, "loss": 0.0269, "lr": 8.304245527866156e-07, "epoch": 18.51956624233852, "percentage": 92.6, "elapsed_time": "2:22:09", "remaining_time": "0:11:21", "throughput": 4631.78, "total_tokens": 39508608} +{"current_steps": 39285, "total_steps": 42420, "loss": 0.3023, "lr": 8.277975112134878e-07, "epoch": 18.52192362093352, "percentage": 92.61, "elapsed_time": "2:22:10", "remaining_time": "0:11:20", "throughput": 4631.77, "total_tokens": 39512928} +{"current_steps": 39290, "total_steps": 42420, "loss": 0.1166, "lr": 8.251745615838192e-07, "epoch": 18.524280999528525, "percentage": 92.62, "elapsed_time": "2:22:11", "remaining_time": "0:11:19", "throughput": 4631.78, "total_tokens": 39517344} +{"current_steps": 39295, "total_steps": 42420, "loss": 0.2468, "lr": 8.225557043416349e-07, "epoch": 18.526638378123526, "percentage": 92.63, "elapsed_time": "2:22:12", "remaining_time": "0:11:18", "throughput": 4631.8, "total_tokens": 39522656} +{"current_steps": 39300, "total_steps": 42420, "loss": 0.1155, "lr": 8.199409399302582e-07, "epoch": 18.52899575671853, "percentage": 92.64, "elapsed_time": "2:22:13", "remaining_time": "0:11:17", "throughput": 4631.81, "total_tokens": 39527488} +{"current_steps": 39305, "total_steps": 42420, "loss": 0.1965, "lr": 8.17330268792324e-07, "epoch": 18.53135313531353, "percentage": 92.66, "elapsed_time": "2:22:14", "remaining_time": "0:11:16", "throughput": 4631.81, "total_tokens": 39531456} +{"current_steps": 39310, "total_steps": 42420, "loss": 0.043, "lr": 8.147236913697787e-07, "epoch": 18.533710513908535, "percentage": 92.67, "elapsed_time": "2:22:15", "remaining_time": "0:11:15", "throughput": 4631.83, "total_tokens": 39535968} +{"current_steps": 39315, "total_steps": 42420, "loss": 0.0509, "lr": 8.121212081038721e-07, "epoch": 18.536067892503535, "percentage": 92.68, "elapsed_time": "2:22:16", "remaining_time": "0:11:14", "throughput": 4631.87, "total_tokens": 39541376} +{"current_steps": 39320, "total_steps": 42420, "loss": 0.1655, "lr": 8.095228194351545e-07, "epoch": 18.53842527109854, "percentage": 92.69, "elapsed_time": "2:22:17", "remaining_time": "0:11:13", "throughput": 4631.91, "total_tokens": 39546368} +{"current_steps": 39325, "total_steps": 42420, "loss": 0.0174, "lr": 8.069285258034936e-07, "epoch": 18.54078264969354, "percentage": 92.7, "elapsed_time": "2:22:18", "remaining_time": "0:11:12", "throughput": 4631.93, "total_tokens": 39551200} +{"current_steps": 39330, "total_steps": 42420, "loss": 0.1291, "lr": 8.043383276480603e-07, "epoch": 18.543140028288544, "percentage": 92.72, "elapsed_time": "2:22:19", "remaining_time": "0:11:10", "throughput": 4631.96, "total_tokens": 39556736} +{"current_steps": 39335, "total_steps": 42420, "loss": 0.1175, "lr": 8.017522254073289e-07, "epoch": 18.545497406883545, "percentage": 92.73, "elapsed_time": "2:22:21", "remaining_time": "0:11:09", "throughput": 4631.96, "total_tokens": 39561856} +{"current_steps": 39340, "total_steps": 42420, "loss": 0.1836, "lr": 7.991702195190854e-07, "epoch": 18.54785478547855, "percentage": 92.74, "elapsed_time": "2:22:22", "remaining_time": "0:11:08", "throughput": 4631.98, "total_tokens": 39566816} +{"current_steps": 39345, "total_steps": 42420, "loss": 0.1644, "lr": 7.965923104204131e-07, "epoch": 18.55021216407355, "percentage": 92.75, "elapsed_time": "2:22:23", "remaining_time": "0:11:07", "throughput": 4632.0, "total_tokens": 39571808} +{"current_steps": 39350, "total_steps": 42420, "loss": 0.0728, "lr": 7.940184985477133e-07, "epoch": 18.552569542668554, "percentage": 92.76, "elapsed_time": "2:22:24", "remaining_time": "0:11:06", "throughput": 4632.02, "total_tokens": 39576544} +{"current_steps": 39355, "total_steps": 42420, "loss": 0.0288, "lr": 7.914487843366874e-07, "epoch": 18.554926921263554, "percentage": 92.77, "elapsed_time": "2:22:25", "remaining_time": "0:11:05", "throughput": 4632.02, "total_tokens": 39580864} +{"current_steps": 39360, "total_steps": 42420, "loss": 0.0355, "lr": 7.88883168222343e-07, "epoch": 18.55728429985856, "percentage": 92.79, "elapsed_time": "2:22:26", "remaining_time": "0:11:04", "throughput": 4632.05, "total_tokens": 39585824} +{"current_steps": 39365, "total_steps": 42420, "loss": 0.1453, "lr": 7.863216506389964e-07, "epoch": 18.55964167845356, "percentage": 92.8, "elapsed_time": "2:22:27", "remaining_time": "0:11:03", "throughput": 4632.06, "total_tokens": 39591072} +{"current_steps": 39370, "total_steps": 42420, "loss": 0.1876, "lr": 7.837642320202732e-07, "epoch": 18.561999057048563, "percentage": 92.81, "elapsed_time": "2:22:28", "remaining_time": "0:11:02", "throughput": 4632.06, "total_tokens": 39595616} +{"current_steps": 39375, "total_steps": 42420, "loss": 0.188, "lr": 7.812109127990935e-07, "epoch": 18.564356435643564, "percentage": 92.82, "elapsed_time": "2:22:29", "remaining_time": "0:11:01", "throughput": 4632.08, "total_tokens": 39600608} +{"current_steps": 39380, "total_steps": 42420, "loss": 0.1367, "lr": 7.786616934076979e-07, "epoch": 18.566713814238568, "percentage": 92.83, "elapsed_time": "2:22:30", "remaining_time": "0:11:00", "throughput": 4632.08, "total_tokens": 39604832} +{"current_steps": 39385, "total_steps": 42420, "loss": 0.2744, "lr": 7.761165742776217e-07, "epoch": 18.56907119283357, "percentage": 92.85, "elapsed_time": "2:22:31", "remaining_time": "0:10:58", "throughput": 4632.08, "total_tokens": 39610144} +{"current_steps": 39390, "total_steps": 42420, "loss": 0.1263, "lr": 7.73575555839709e-07, "epoch": 18.571428571428573, "percentage": 92.86, "elapsed_time": "2:22:32", "remaining_time": "0:10:57", "throughput": 4632.09, "total_tokens": 39615168} +{"current_steps": 39395, "total_steps": 42420, "loss": 0.1583, "lr": 7.710386385241158e-07, "epoch": 18.573785950023574, "percentage": 92.87, "elapsed_time": "2:22:33", "remaining_time": "0:10:56", "throughput": 4632.09, "total_tokens": 39619392} +{"current_steps": 39400, "total_steps": 42420, "loss": 0.0661, "lr": 7.685058227602959e-07, "epoch": 18.576143328618578, "percentage": 92.88, "elapsed_time": "2:22:34", "remaining_time": "0:10:55", "throughput": 4632.1, "total_tokens": 39623392} +{"current_steps": 39405, "total_steps": 42420, "loss": 0.1325, "lr": 7.659771089770118e-07, "epoch": 18.57850070721358, "percentage": 92.89, "elapsed_time": "2:22:35", "remaining_time": "0:10:54", "throughput": 4632.08, "total_tokens": 39627456} +{"current_steps": 39410, "total_steps": 42420, "loss": 0.1456, "lr": 7.634524976023349e-07, "epoch": 18.580858085808583, "percentage": 92.9, "elapsed_time": "2:22:35", "remaining_time": "0:10:53", "throughput": 4632.09, "total_tokens": 39632096} +{"current_steps": 39415, "total_steps": 42420, "loss": 0.1843, "lr": 7.609319890636346e-07, "epoch": 18.583215464403583, "percentage": 92.92, "elapsed_time": "2:22:37", "remaining_time": "0:10:52", "throughput": 4632.12, "total_tokens": 39637632} +{"current_steps": 39420, "total_steps": 42420, "loss": 0.1558, "lr": 7.584155837875973e-07, "epoch": 18.585572842998587, "percentage": 92.93, "elapsed_time": "2:22:38", "remaining_time": "0:10:51", "throughput": 4632.14, "total_tokens": 39642432} +{"current_steps": 39425, "total_steps": 42420, "loss": 0.0229, "lr": 7.559032822001988e-07, "epoch": 18.587930221593588, "percentage": 92.94, "elapsed_time": "2:22:39", "remaining_time": "0:10:50", "throughput": 4632.16, "total_tokens": 39647808} +{"current_steps": 39430, "total_steps": 42420, "loss": 0.1226, "lr": 7.533950847267351e-07, "epoch": 18.59028760018859, "percentage": 92.95, "elapsed_time": "2:22:40", "remaining_time": "0:10:49", "throughput": 4632.19, "total_tokens": 39652864} +{"current_steps": 39435, "total_steps": 42420, "loss": 0.0955, "lr": 7.508909917917972e-07, "epoch": 18.592644978783593, "percentage": 92.96, "elapsed_time": "2:22:41", "remaining_time": "0:10:48", "throughput": 4632.22, "total_tokens": 39658560} +{"current_steps": 39440, "total_steps": 42420, "loss": 0.0357, "lr": 7.483910038192904e-07, "epoch": 18.595002357378593, "percentage": 92.98, "elapsed_time": "2:22:42", "remaining_time": "0:10:46", "throughput": 4632.24, "total_tokens": 39663168} +{"current_steps": 39445, "total_steps": 42420, "loss": 0.1306, "lr": 7.458951212324178e-07, "epoch": 18.597359735973598, "percentage": 92.99, "elapsed_time": "2:22:43", "remaining_time": "0:10:45", "throughput": 4632.25, "total_tokens": 39667712} +{"current_steps": 39450, "total_steps": 42420, "loss": 0.1479, "lr": 7.434033444536886e-07, "epoch": 18.599717114568598, "percentage": 93.0, "elapsed_time": "2:22:44", "remaining_time": "0:10:44", "throughput": 4632.27, "total_tokens": 39672256} +{"current_steps": 39455, "total_steps": 42420, "loss": 0.0728, "lr": 7.409156739049183e-07, "epoch": 18.602074493163602, "percentage": 93.01, "elapsed_time": "2:22:45", "remaining_time": "0:10:43", "throughput": 4632.28, "total_tokens": 39676608} +{"current_steps": 39460, "total_steps": 42420, "loss": 0.0778, "lr": 7.384321100072339e-07, "epoch": 18.604431871758603, "percentage": 93.02, "elapsed_time": "2:22:46", "remaining_time": "0:10:42", "throughput": 4632.3, "total_tokens": 39680992} +{"current_steps": 39465, "total_steps": 42420, "loss": 0.1383, "lr": 7.359526531810518e-07, "epoch": 18.606789250353607, "percentage": 93.03, "elapsed_time": "2:22:47", "remaining_time": "0:10:41", "throughput": 4632.31, "total_tokens": 39686048} +{"current_steps": 39470, "total_steps": 42420, "loss": 0.2182, "lr": 7.334773038461056e-07, "epoch": 18.609146628948608, "percentage": 93.05, "elapsed_time": "2:22:48", "remaining_time": "0:10:40", "throughput": 4632.35, "total_tokens": 39691584} +{"current_steps": 39475, "total_steps": 42420, "loss": 0.0293, "lr": 7.310060624214271e-07, "epoch": 18.611504007543612, "percentage": 93.06, "elapsed_time": "2:22:49", "remaining_time": "0:10:39", "throughput": 4632.38, "total_tokens": 39697504} +{"current_steps": 39480, "total_steps": 42420, "loss": 0.0858, "lr": 7.285389293253619e-07, "epoch": 18.613861386138613, "percentage": 93.07, "elapsed_time": "2:22:50", "remaining_time": "0:10:38", "throughput": 4632.42, "total_tokens": 39702816} +{"current_steps": 39485, "total_steps": 42420, "loss": 0.0385, "lr": 7.260759049755484e-07, "epoch": 18.616218764733617, "percentage": 93.08, "elapsed_time": "2:22:52", "remaining_time": "0:10:37", "throughput": 4632.48, "total_tokens": 39709792} +{"current_steps": 39490, "total_steps": 42420, "loss": 0.141, "lr": 7.236169897889361e-07, "epoch": 18.618576143328617, "percentage": 93.09, "elapsed_time": "2:22:53", "remaining_time": "0:10:36", "throughput": 4632.49, "total_tokens": 39714368} +{"current_steps": 39495, "total_steps": 42420, "loss": 0.0161, "lr": 7.211621841817784e-07, "epoch": 18.62093352192362, "percentage": 93.1, "elapsed_time": "2:22:54", "remaining_time": "0:10:34", "throughput": 4632.51, "total_tokens": 39719392} +{"current_steps": 39500, "total_steps": 42420, "loss": 0.1927, "lr": 7.187114885696345e-07, "epoch": 18.623290900518622, "percentage": 93.12, "elapsed_time": "2:22:55", "remaining_time": "0:10:33", "throughput": 4632.5, "total_tokens": 39723904} +{"current_steps": 39505, "total_steps": 42420, "loss": 0.0882, "lr": 7.162649033673585e-07, "epoch": 18.625648279113626, "percentage": 93.13, "elapsed_time": "2:22:55", "remaining_time": "0:10:32", "throughput": 4632.51, "total_tokens": 39728000} +{"current_steps": 39510, "total_steps": 42420, "loss": 0.0375, "lr": 7.138224289891221e-07, "epoch": 18.628005657708627, "percentage": 93.14, "elapsed_time": "2:22:56", "remaining_time": "0:10:31", "throughput": 4632.52, "total_tokens": 39732608} +{"current_steps": 39515, "total_steps": 42420, "loss": 0.0125, "lr": 7.113840658483917e-07, "epoch": 18.63036303630363, "percentage": 93.15, "elapsed_time": "2:22:58", "remaining_time": "0:10:30", "throughput": 4632.56, "total_tokens": 39738144} +{"current_steps": 39520, "total_steps": 42420, "loss": 0.0554, "lr": 7.089498143579426e-07, "epoch": 18.632720414898632, "percentage": 93.16, "elapsed_time": "2:22:59", "remaining_time": "0:10:29", "throughput": 4632.58, "total_tokens": 39743456} +{"current_steps": 39525, "total_steps": 42420, "loss": 0.2226, "lr": 7.065196749298508e-07, "epoch": 18.635077793493636, "percentage": 93.18, "elapsed_time": "2:23:00", "remaining_time": "0:10:28", "throughput": 4632.59, "total_tokens": 39749120} +{"current_steps": 39530, "total_steps": 42420, "loss": 0.1352, "lr": 7.040936479754984e-07, "epoch": 18.637435172088637, "percentage": 93.19, "elapsed_time": "2:23:01", "remaining_time": "0:10:27", "throughput": 4632.6, "total_tokens": 39753504} +{"current_steps": 39535, "total_steps": 42420, "loss": 0.02, "lr": 7.016717339055706e-07, "epoch": 18.63979255068364, "percentage": 93.2, "elapsed_time": "2:23:02", "remaining_time": "0:10:26", "throughput": 4632.62, "total_tokens": 39757984} +{"current_steps": 39540, "total_steps": 42420, "loss": 0.0798, "lr": 6.99253933130059e-07, "epoch": 18.64214992927864, "percentage": 93.21, "elapsed_time": "2:23:03", "remaining_time": "0:10:25", "throughput": 4632.65, "total_tokens": 39763136} +{"current_steps": 39545, "total_steps": 42420, "loss": 0.0883, "lr": 6.968402460582502e-07, "epoch": 18.644507307873646, "percentage": 93.22, "elapsed_time": "2:23:04", "remaining_time": "0:10:24", "throughput": 4632.67, "total_tokens": 39768416} +{"current_steps": 39550, "total_steps": 42420, "loss": 0.0637, "lr": 6.944306730987421e-07, "epoch": 18.646864686468646, "percentage": 93.23, "elapsed_time": "2:23:05", "remaining_time": "0:10:23", "throughput": 4632.71, "total_tokens": 39774208} +{"current_steps": 39555, "total_steps": 42420, "loss": 0.016, "lr": 6.920252146594364e-07, "epoch": 18.64922206506365, "percentage": 93.25, "elapsed_time": "2:23:06", "remaining_time": "0:10:21", "throughput": 4632.66, "total_tokens": 39778080} +{"current_steps": 39560, "total_steps": 42420, "loss": 0.1284, "lr": 6.896238711475378e-07, "epoch": 18.65157944365865, "percentage": 93.26, "elapsed_time": "2:23:07", "remaining_time": "0:10:20", "throughput": 4632.71, "total_tokens": 39784608} +{"current_steps": 39565, "total_steps": 42420, "loss": 0.1575, "lr": 6.872266429695518e-07, "epoch": 18.653936822253655, "percentage": 93.27, "elapsed_time": "2:23:08", "remaining_time": "0:10:19", "throughput": 4632.72, "total_tokens": 39788736} +{"current_steps": 39570, "total_steps": 42420, "loss": 0.0156, "lr": 6.84833530531287e-07, "epoch": 18.656294200848656, "percentage": 93.28, "elapsed_time": "2:23:09", "remaining_time": "0:10:18", "throughput": 4632.74, "total_tokens": 39793600} +{"current_steps": 39575, "total_steps": 42420, "loss": 0.1292, "lr": 6.824445342378583e-07, "epoch": 18.65865157944366, "percentage": 93.29, "elapsed_time": "2:23:10", "remaining_time": "0:10:17", "throughput": 4632.79, "total_tokens": 39799104} +{"current_steps": 39580, "total_steps": 42420, "loss": 0.1113, "lr": 6.800596544936783e-07, "epoch": 18.66100895803866, "percentage": 93.31, "elapsed_time": "2:23:11", "remaining_time": "0:10:16", "throughput": 4632.81, "total_tokens": 39803872} +{"current_steps": 39585, "total_steps": 42420, "loss": 0.3191, "lr": 6.776788917024713e-07, "epoch": 18.663366336633665, "percentage": 93.32, "elapsed_time": "2:23:12", "remaining_time": "0:10:15", "throughput": 4632.85, "total_tokens": 39809504} +{"current_steps": 39590, "total_steps": 42420, "loss": 0.0296, "lr": 6.753022462672592e-07, "epoch": 18.665723715228665, "percentage": 93.33, "elapsed_time": "2:23:13", "remaining_time": "0:10:14", "throughput": 4632.85, "total_tokens": 39813664} +{"current_steps": 39595, "total_steps": 42420, "loss": 0.0908, "lr": 6.729297185903677e-07, "epoch": 18.66808109382367, "percentage": 93.34, "elapsed_time": "2:23:15", "remaining_time": "0:10:13", "throughput": 4632.9, "total_tokens": 39821312} +{"current_steps": 39600, "total_steps": 42420, "loss": 0.0323, "lr": 6.705613090734225e-07, "epoch": 18.67043847241867, "percentage": 93.35, "elapsed_time": "2:23:16", "remaining_time": "0:10:12", "throughput": 4632.9, "total_tokens": 39825856} +{"current_steps": 39605, "total_steps": 42420, "loss": 0.0556, "lr": 6.681970181173585e-07, "epoch": 18.672795851013674, "percentage": 93.36, "elapsed_time": "2:23:17", "remaining_time": "0:10:11", "throughput": 4632.92, "total_tokens": 39830912} +{"current_steps": 39610, "total_steps": 42420, "loss": 0.0655, "lr": 6.658368461224085e-07, "epoch": 18.675153229608675, "percentage": 93.38, "elapsed_time": "2:23:18", "remaining_time": "0:10:09", "throughput": 4632.95, "total_tokens": 39835936} +{"current_steps": 39615, "total_steps": 42420, "loss": 0.1309, "lr": 6.634807934881082e-07, "epoch": 18.677510608203676, "percentage": 93.39, "elapsed_time": "2:23:19", "remaining_time": "0:10:08", "throughput": 4632.98, "total_tokens": 39841440} +{"current_steps": 39620, "total_steps": 42420, "loss": 0.1338, "lr": 6.611288606132998e-07, "epoch": 18.67986798679868, "percentage": 93.4, "elapsed_time": "2:23:20", "remaining_time": "0:10:07", "throughput": 4633.0, "total_tokens": 39845760} +{"current_steps": 39625, "total_steps": 42420, "loss": 0.3115, "lr": 6.587810478961259e-07, "epoch": 18.68222536539368, "percentage": 93.41, "elapsed_time": "2:23:21", "remaining_time": "0:10:06", "throughput": 4633.03, "total_tokens": 39850848} +{"current_steps": 39630, "total_steps": 42420, "loss": 0.0286, "lr": 6.564373557340326e-07, "epoch": 18.684582743988685, "percentage": 93.42, "elapsed_time": "2:23:22", "remaining_time": "0:10:05", "throughput": 4633.07, "total_tokens": 39856384} +{"current_steps": 39635, "total_steps": 42420, "loss": 0.1787, "lr": 6.540977845237606e-07, "epoch": 18.686940122583685, "percentage": 93.43, "elapsed_time": "2:23:23", "remaining_time": "0:10:04", "throughput": 4633.09, "total_tokens": 39860864} +{"current_steps": 39640, "total_steps": 42420, "loss": 0.0826, "lr": 6.517623346613654e-07, "epoch": 18.68929750117869, "percentage": 93.45, "elapsed_time": "2:23:24", "remaining_time": "0:10:03", "throughput": 4633.1, "total_tokens": 39866080} +{"current_steps": 39645, "total_steps": 42420, "loss": 0.245, "lr": 6.494310065421977e-07, "epoch": 18.69165487977369, "percentage": 93.46, "elapsed_time": "2:23:25", "remaining_time": "0:10:02", "throughput": 4633.15, "total_tokens": 39871936} +{"current_steps": 39650, "total_steps": 42420, "loss": 0.064, "lr": 6.471038005609109e-07, "epoch": 18.694012258368694, "percentage": 93.47, "elapsed_time": "2:23:26", "remaining_time": "0:10:01", "throughput": 4633.17, "total_tokens": 39876512} +{"current_steps": 39655, "total_steps": 42420, "loss": 0.0385, "lr": 6.447807171114622e-07, "epoch": 18.696369636963695, "percentage": 93.48, "elapsed_time": "2:23:27", "remaining_time": "0:10:00", "throughput": 4633.18, "total_tokens": 39880992} +{"current_steps": 39660, "total_steps": 42420, "loss": 0.0616, "lr": 6.424617565871094e-07, "epoch": 18.6987270155587, "percentage": 93.49, "elapsed_time": "2:23:28", "remaining_time": "0:09:59", "throughput": 4633.2, "total_tokens": 39886208} +{"current_steps": 39665, "total_steps": 42420, "loss": 0.1621, "lr": 6.401469193804133e-07, "epoch": 18.7010843941537, "percentage": 93.51, "elapsed_time": "2:23:29", "remaining_time": "0:09:58", "throughput": 4633.26, "total_tokens": 39892192} +{"current_steps": 39670, "total_steps": 42420, "loss": 0.0417, "lr": 6.37836205883241e-07, "epoch": 18.703441772748704, "percentage": 93.52, "elapsed_time": "2:23:31", "remaining_time": "0:09:56", "throughput": 4633.3, "total_tokens": 39897824} +{"current_steps": 39675, "total_steps": 42420, "loss": 0.2497, "lr": 6.355296164867491e-07, "epoch": 18.705799151343705, "percentage": 93.53, "elapsed_time": "2:23:32", "remaining_time": "0:09:55", "throughput": 4633.32, "total_tokens": 39903456} +{"current_steps": 39680, "total_steps": 42420, "loss": 0.22, "lr": 6.332271515814087e-07, "epoch": 18.70815652993871, "percentage": 93.54, "elapsed_time": "2:23:33", "remaining_time": "0:09:54", "throughput": 4633.34, "total_tokens": 39908160} +{"current_steps": 39685, "total_steps": 42420, "loss": 0.1645, "lr": 6.309288115569884e-07, "epoch": 18.71051390853371, "percentage": 93.55, "elapsed_time": "2:23:34", "remaining_time": "0:09:53", "throughput": 4633.37, "total_tokens": 39913696} +{"current_steps": 39690, "total_steps": 42420, "loss": 0.1978, "lr": 6.286345968025548e-07, "epoch": 18.712871287128714, "percentage": 93.56, "elapsed_time": "2:23:35", "remaining_time": "0:09:52", "throughput": 4633.37, "total_tokens": 39918432} +{"current_steps": 39695, "total_steps": 42420, "loss": 0.043, "lr": 6.263445077064833e-07, "epoch": 18.715228665723714, "percentage": 93.58, "elapsed_time": "2:23:36", "remaining_time": "0:09:51", "throughput": 4633.41, "total_tokens": 39923872} +{"current_steps": 39700, "total_steps": 42420, "loss": 0.0253, "lr": 6.240585446564446e-07, "epoch": 18.71758604431872, "percentage": 93.59, "elapsed_time": "2:23:37", "remaining_time": "0:09:50", "throughput": 4633.44, "total_tokens": 39928928} +{"current_steps": 39705, "total_steps": 42420, "loss": 0.0518, "lr": 6.217767080394149e-07, "epoch": 18.71994342291372, "percentage": 93.6, "elapsed_time": "2:23:38", "remaining_time": "0:09:49", "throughput": 4633.45, "total_tokens": 39933536} +{"current_steps": 39710, "total_steps": 42420, "loss": 0.2999, "lr": 6.194989982416715e-07, "epoch": 18.722300801508723, "percentage": 93.61, "elapsed_time": "2:23:39", "remaining_time": "0:09:48", "throughput": 4633.46, "total_tokens": 39938048} +{"current_steps": 39715, "total_steps": 42420, "loss": 0.0915, "lr": 6.172254156487894e-07, "epoch": 18.724658180103724, "percentage": 93.62, "elapsed_time": "2:23:40", "remaining_time": "0:09:47", "throughput": 4633.48, "total_tokens": 39942752} +{"current_steps": 39720, "total_steps": 42420, "loss": 0.1914, "lr": 6.149559606456468e-07, "epoch": 18.727015558698728, "percentage": 93.64, "elapsed_time": "2:23:41", "remaining_time": "0:09:46", "throughput": 4633.47, "total_tokens": 39946784} +{"current_steps": 39725, "total_steps": 42420, "loss": 0.0627, "lr": 6.126906336164279e-07, "epoch": 18.72937293729373, "percentage": 93.65, "elapsed_time": "2:23:42", "remaining_time": "0:09:44", "throughput": 4633.5, "total_tokens": 39952384} +{"current_steps": 39730, "total_steps": 42420, "loss": 0.0264, "lr": 6.104294349446094e-07, "epoch": 18.731730315888733, "percentage": 93.66, "elapsed_time": "2:23:43", "remaining_time": "0:09:43", "throughput": 4633.53, "total_tokens": 39956672} +{"current_steps": 39735, "total_steps": 42420, "loss": 0.1414, "lr": 6.081723650129767e-07, "epoch": 18.734087694483733, "percentage": 93.67, "elapsed_time": "2:23:44", "remaining_time": "0:09:42", "throughput": 4633.57, "total_tokens": 39962368} +{"current_steps": 39740, "total_steps": 42420, "loss": 0.2651, "lr": 6.059194242036131e-07, "epoch": 18.736445073078738, "percentage": 93.68, "elapsed_time": "2:23:45", "remaining_time": "0:09:41", "throughput": 4633.64, "total_tokens": 39969728} +{"current_steps": 39745, "total_steps": 42420, "loss": 0.1204, "lr": 6.036706128979053e-07, "epoch": 18.738802451673738, "percentage": 93.69, "elapsed_time": "2:23:46", "remaining_time": "0:09:40", "throughput": 4633.67, "total_tokens": 39974656} +{"current_steps": 39750, "total_steps": 42420, "loss": 0.2363, "lr": 6.014259314765375e-07, "epoch": 18.741159830268742, "percentage": 93.71, "elapsed_time": "2:23:47", "remaining_time": "0:09:39", "throughput": 4633.67, "total_tokens": 39978976} +{"current_steps": 39755, "total_steps": 42420, "loss": 0.1488, "lr": 5.991853803194919e-07, "epoch": 18.743517208863743, "percentage": 93.72, "elapsed_time": "2:23:48", "remaining_time": "0:09:38", "throughput": 4633.69, "total_tokens": 39984000} +{"current_steps": 39760, "total_steps": 42420, "loss": 0.0711, "lr": 5.969489598060596e-07, "epoch": 18.745874587458747, "percentage": 93.73, "elapsed_time": "2:23:49", "remaining_time": "0:09:37", "throughput": 4633.68, "total_tokens": 39988512} +{"current_steps": 39765, "total_steps": 42420, "loss": 0.2039, "lr": 5.947166703148293e-07, "epoch": 18.748231966053748, "percentage": 93.74, "elapsed_time": "2:23:50", "remaining_time": "0:09:36", "throughput": 4633.69, "total_tokens": 39992832} +{"current_steps": 39770, "total_steps": 42420, "loss": 0.0441, "lr": 5.924885122236851e-07, "epoch": 18.750589344648752, "percentage": 93.75, "elapsed_time": "2:23:52", "remaining_time": "0:09:35", "throughput": 4633.74, "total_tokens": 39999072} +{"current_steps": 39775, "total_steps": 42420, "loss": 0.0406, "lr": 5.90264485909825e-07, "epoch": 18.752946723243753, "percentage": 93.76, "elapsed_time": "2:23:53", "remaining_time": "0:09:34", "throughput": 4633.78, "total_tokens": 40004352} +{"current_steps": 39780, "total_steps": 42420, "loss": 0.1138, "lr": 5.880445917497284e-07, "epoch": 18.755304101838757, "percentage": 93.78, "elapsed_time": "2:23:54", "remaining_time": "0:09:33", "throughput": 4633.82, "total_tokens": 40010432} +{"current_steps": 39785, "total_steps": 42420, "loss": 0.2268, "lr": 5.858288301191949e-07, "epoch": 18.757661480433757, "percentage": 93.79, "elapsed_time": "2:23:55", "remaining_time": "0:09:31", "throughput": 4633.68, "total_tokens": 40015296} +{"current_steps": 39790, "total_steps": 42420, "loss": 0.0311, "lr": 5.836172013933133e-07, "epoch": 18.76001885902876, "percentage": 93.8, "elapsed_time": "2:23:56", "remaining_time": "0:09:30", "throughput": 4633.72, "total_tokens": 40020960} +{"current_steps": 39795, "total_steps": 42420, "loss": 0.2128, "lr": 5.814097059464702e-07, "epoch": 18.762376237623762, "percentage": 93.81, "elapsed_time": "2:23:57", "remaining_time": "0:09:29", "throughput": 4633.75, "total_tokens": 40026272} +{"current_steps": 39800, "total_steps": 42420, "loss": 0.0159, "lr": 5.792063441523609e-07, "epoch": 18.764733616218766, "percentage": 93.82, "elapsed_time": "2:23:59", "remaining_time": "0:09:28", "throughput": 4633.77, "total_tokens": 40031232} +{"current_steps": 39805, "total_steps": 42420, "loss": 0.1002, "lr": 5.770071163839763e-07, "epoch": 18.767090994813767, "percentage": 93.84, "elapsed_time": "2:24:00", "remaining_time": "0:09:27", "throughput": 4633.79, "total_tokens": 40036032} +{"current_steps": 39810, "total_steps": 42420, "loss": 0.0968, "lr": 5.748120230136045e-07, "epoch": 18.76944837340877, "percentage": 93.85, "elapsed_time": "2:24:01", "remaining_time": "0:09:26", "throughput": 4633.81, "total_tokens": 40040832} +{"current_steps": 39815, "total_steps": 42420, "loss": 0.0528, "lr": 5.726210644128427e-07, "epoch": 18.77180575200377, "percentage": 93.86, "elapsed_time": "2:24:02", "remaining_time": "0:09:25", "throughput": 4633.84, "total_tokens": 40045664} +{"current_steps": 39820, "total_steps": 42420, "loss": 0.088, "lr": 5.704342409525831e-07, "epoch": 18.774163130598772, "percentage": 93.87, "elapsed_time": "2:24:03", "remaining_time": "0:09:24", "throughput": 4633.9, "total_tokens": 40053024} +{"current_steps": 39825, "total_steps": 42420, "loss": 0.1828, "lr": 5.682515530030158e-07, "epoch": 18.776520509193777, "percentage": 93.88, "elapsed_time": "2:24:04", "remaining_time": "0:09:23", "throughput": 4633.9, "total_tokens": 40056928} +{"current_steps": 39830, "total_steps": 42420, "loss": 0.053, "lr": 5.660730009336312e-07, "epoch": 18.778877887788777, "percentage": 93.89, "elapsed_time": "2:24:05", "remaining_time": "0:09:22", "throughput": 4633.92, "total_tokens": 40061888} +{"current_steps": 39835, "total_steps": 42420, "loss": 0.0805, "lr": 5.638985851132234e-07, "epoch": 18.78123526638378, "percentage": 93.91, "elapsed_time": "2:24:06", "remaining_time": "0:09:21", "throughput": 4633.95, "total_tokens": 40067552} +{"current_steps": 39840, "total_steps": 42420, "loss": 0.0252, "lr": 5.617283059098865e-07, "epoch": 18.783592644978782, "percentage": 93.92, "elapsed_time": "2:24:07", "remaining_time": "0:09:20", "throughput": 4633.99, "total_tokens": 40073280} +{"current_steps": 39845, "total_steps": 42420, "loss": 0.2854, "lr": 5.595621636910075e-07, "epoch": 18.785950023573786, "percentage": 93.93, "elapsed_time": "2:24:08", "remaining_time": "0:09:18", "throughput": 4634.04, "total_tokens": 40079584} +{"current_steps": 39850, "total_steps": 42420, "loss": 0.0205, "lr": 5.574001588232791e-07, "epoch": 18.788307402168787, "percentage": 93.94, "elapsed_time": "2:24:09", "remaining_time": "0:09:17", "throughput": 4634.05, "total_tokens": 40083936} +{"current_steps": 39855, "total_steps": 42420, "loss": 0.106, "lr": 5.552422916726918e-07, "epoch": 18.79066478076379, "percentage": 93.95, "elapsed_time": "2:24:10", "remaining_time": "0:09:16", "throughput": 4634.04, "total_tokens": 40087936} +{"current_steps": 39860, "total_steps": 42420, "loss": 0.0457, "lr": 5.53088562604534e-07, "epoch": 18.79302215935879, "percentage": 93.97, "elapsed_time": "2:24:11", "remaining_time": "0:09:15", "throughput": 4633.98, "total_tokens": 40091296} +{"current_steps": 39865, "total_steps": 42420, "loss": 0.0172, "lr": 5.509389719834002e-07, "epoch": 18.795379537953796, "percentage": 93.98, "elapsed_time": "2:24:12", "remaining_time": "0:09:14", "throughput": 4634.0, "total_tokens": 40096608} +{"current_steps": 39870, "total_steps": 42420, "loss": 0.1239, "lr": 5.48793520173177e-07, "epoch": 18.797736916548796, "percentage": 93.99, "elapsed_time": "2:24:13", "remaining_time": "0:09:13", "throughput": 4634.04, "total_tokens": 40101824} +{"current_steps": 39875, "total_steps": 42420, "loss": 0.0354, "lr": 5.466522075370517e-07, "epoch": 18.8000942951438, "percentage": 94.0, "elapsed_time": "2:24:14", "remaining_time": "0:09:12", "throughput": 4634.05, "total_tokens": 40106432} +{"current_steps": 39880, "total_steps": 42420, "loss": 0.12, "lr": 5.445150344375178e-07, "epoch": 18.8024516737388, "percentage": 94.01, "elapsed_time": "2:24:16", "remaining_time": "0:09:11", "throughput": 4634.09, "total_tokens": 40113408} +{"current_steps": 39885, "total_steps": 42420, "loss": 0.0595, "lr": 5.42382001236355e-07, "epoch": 18.804809052333805, "percentage": 94.02, "elapsed_time": "2:24:17", "remaining_time": "0:09:10", "throughput": 4634.1, "total_tokens": 40117888} +{"current_steps": 39890, "total_steps": 42420, "loss": 0.1484, "lr": 5.402531082946499e-07, "epoch": 18.807166430928806, "percentage": 94.04, "elapsed_time": "2:24:18", "remaining_time": "0:09:09", "throughput": 4634.14, "total_tokens": 40123360} +{"current_steps": 39895, "total_steps": 42420, "loss": 0.1145, "lr": 5.381283559727945e-07, "epoch": 18.80952380952381, "percentage": 94.05, "elapsed_time": "2:24:19", "remaining_time": "0:09:08", "throughput": 4634.15, "total_tokens": 40128864} +{"current_steps": 39900, "total_steps": 42420, "loss": 0.1867, "lr": 5.360077446304679e-07, "epoch": 18.81188118811881, "percentage": 94.06, "elapsed_time": "2:24:20", "remaining_time": "0:09:06", "throughput": 4634.19, "total_tokens": 40134688} +{"current_steps": 39905, "total_steps": 42420, "loss": 0.0893, "lr": 5.33891274626655e-07, "epoch": 18.814238566713815, "percentage": 94.07, "elapsed_time": "2:24:21", "remaining_time": "0:09:05", "throughput": 4634.21, "total_tokens": 40139488} +{"current_steps": 39910, "total_steps": 42420, "loss": 0.0409, "lr": 5.317789463196388e-07, "epoch": 18.816595945308816, "percentage": 94.08, "elapsed_time": "2:24:22", "remaining_time": "0:09:04", "throughput": 4634.24, "total_tokens": 40144320} +{"current_steps": 39915, "total_steps": 42420, "loss": 0.0211, "lr": 5.296707600669998e-07, "epoch": 18.81895332390382, "percentage": 94.09, "elapsed_time": "2:24:23", "remaining_time": "0:09:03", "throughput": 4634.28, "total_tokens": 40149568} +{"current_steps": 39920, "total_steps": 42420, "loss": 0.2048, "lr": 5.27566716225622e-07, "epoch": 18.82131070249882, "percentage": 94.11, "elapsed_time": "2:24:24", "remaining_time": "0:09:02", "throughput": 4634.28, "total_tokens": 40153632} +{"current_steps": 39925, "total_steps": 42420, "loss": 0.075, "lr": 5.254668151516762e-07, "epoch": 18.823668081093825, "percentage": 94.12, "elapsed_time": "2:24:25", "remaining_time": "0:09:01", "throughput": 4634.33, "total_tokens": 40159296} +{"current_steps": 39930, "total_steps": 42420, "loss": 0.1785, "lr": 5.233710572006473e-07, "epoch": 18.826025459688825, "percentage": 94.13, "elapsed_time": "2:24:26", "remaining_time": "0:09:00", "throughput": 4634.37, "total_tokens": 40165088} +{"current_steps": 39935, "total_steps": 42420, "loss": 0.0268, "lr": 5.212794427273071e-07, "epoch": 18.82838283828383, "percentage": 94.14, "elapsed_time": "2:24:27", "remaining_time": "0:08:59", "throughput": 4634.4, "total_tokens": 40170208} +{"current_steps": 39940, "total_steps": 42420, "loss": 0.137, "lr": 5.191919720857335e-07, "epoch": 18.83074021687883, "percentage": 94.15, "elapsed_time": "2:24:28", "remaining_time": "0:08:58", "throughput": 4634.4, "total_tokens": 40174336} +{"current_steps": 39945, "total_steps": 42420, "loss": 0.1437, "lr": 5.171086456292967e-07, "epoch": 18.833097595473834, "percentage": 94.17, "elapsed_time": "2:24:29", "remaining_time": "0:08:57", "throughput": 4634.42, "total_tokens": 40179136} +{"current_steps": 39950, "total_steps": 42420, "loss": 0.0554, "lr": 5.150294637106728e-07, "epoch": 18.835454974068835, "percentage": 94.18, "elapsed_time": "2:24:30", "remaining_time": "0:08:56", "throughput": 4634.44, "total_tokens": 40184352} +{"current_steps": 39955, "total_steps": 42420, "loss": 0.0812, "lr": 5.129544266818276e-07, "epoch": 18.83781235266384, "percentage": 94.19, "elapsed_time": "2:24:31", "remaining_time": "0:08:55", "throughput": 4634.47, "total_tokens": 40189408} +{"current_steps": 39960, "total_steps": 42420, "loss": 0.1997, "lr": 5.108835348940355e-07, "epoch": 18.84016973125884, "percentage": 94.2, "elapsed_time": "2:24:32", "remaining_time": "0:08:53", "throughput": 4634.5, "total_tokens": 40194400} +{"current_steps": 39965, "total_steps": 42420, "loss": 0.0416, "lr": 5.088167886978579e-07, "epoch": 18.842527109853844, "percentage": 94.21, "elapsed_time": "2:24:33", "remaining_time": "0:08:52", "throughput": 4634.52, "total_tokens": 40199488} +{"current_steps": 39970, "total_steps": 42420, "loss": 0.1078, "lr": 5.067541884431593e-07, "epoch": 18.844884488448844, "percentage": 94.22, "elapsed_time": "2:24:35", "remaining_time": "0:08:51", "throughput": 4634.56, "total_tokens": 40205120} +{"current_steps": 39975, "total_steps": 42420, "loss": 0.1288, "lr": 5.046957344791048e-07, "epoch": 18.84724186704385, "percentage": 94.24, "elapsed_time": "2:24:36", "remaining_time": "0:08:50", "throughput": 4634.6, "total_tokens": 40211712} +{"current_steps": 39980, "total_steps": 42420, "loss": 0.0691, "lr": 5.026414271541574e-07, "epoch": 18.84959924563885, "percentage": 94.25, "elapsed_time": "2:24:37", "remaining_time": "0:08:49", "throughput": 4634.63, "total_tokens": 40217152} +{"current_steps": 39985, "total_steps": 42420, "loss": 0.1671, "lr": 5.005912668160723e-07, "epoch": 18.851956624233853, "percentage": 94.26, "elapsed_time": "2:24:38", "remaining_time": "0:08:48", "throughput": 4634.65, "total_tokens": 40221856} +{"current_steps": 39990, "total_steps": 42420, "loss": 0.1195, "lr": 4.985452538119134e-07, "epoch": 18.854314002828854, "percentage": 94.27, "elapsed_time": "2:24:39", "remaining_time": "0:08:47", "throughput": 4634.69, "total_tokens": 40227232} +{"current_steps": 39995, "total_steps": 42420, "loss": 0.0588, "lr": 4.965033884880288e-07, "epoch": 18.85667138142386, "percentage": 94.28, "elapsed_time": "2:24:40", "remaining_time": "0:08:46", "throughput": 4634.71, "total_tokens": 40231648} +{"current_steps": 40000, "total_steps": 42420, "loss": 0.0151, "lr": 4.94465671190078e-07, "epoch": 18.85902876001886, "percentage": 94.3, "elapsed_time": "2:24:41", "remaining_time": "0:08:45", "throughput": 4634.72, "total_tokens": 40236128} +{"current_steps": 40005, "total_steps": 42420, "loss": 0.0895, "lr": 4.924321022630074e-07, "epoch": 18.861386138613863, "percentage": 94.31, "elapsed_time": "2:24:42", "remaining_time": "0:08:44", "throughput": 4634.74, "total_tokens": 40241088} +{"current_steps": 40010, "total_steps": 42420, "loss": 0.0867, "lr": 4.904026820510665e-07, "epoch": 18.863743517208864, "percentage": 94.32, "elapsed_time": "2:24:43", "remaining_time": "0:08:43", "throughput": 4634.76, "total_tokens": 40245888} +{"current_steps": 40015, "total_steps": 42420, "loss": 0.0626, "lr": 4.883774108978029e-07, "epoch": 18.866100895803868, "percentage": 94.33, "elapsed_time": "2:24:44", "remaining_time": "0:08:41", "throughput": 4634.79, "total_tokens": 40250944} +{"current_steps": 40020, "total_steps": 42420, "loss": 0.156, "lr": 4.863562891460588e-07, "epoch": 18.86845827439887, "percentage": 94.34, "elapsed_time": "2:24:45", "remaining_time": "0:08:40", "throughput": 4634.81, "total_tokens": 40255968} +{"current_steps": 40025, "total_steps": 42420, "loss": 0.0963, "lr": 4.843393171379773e-07, "epoch": 18.87081565299387, "percentage": 94.35, "elapsed_time": "2:24:46", "remaining_time": "0:08:39", "throughput": 4634.83, "total_tokens": 40261216} +{"current_steps": 40030, "total_steps": 42420, "loss": 0.0298, "lr": 4.823264952149964e-07, "epoch": 18.873173031588873, "percentage": 94.37, "elapsed_time": "2:24:47", "remaining_time": "0:08:38", "throughput": 4634.87, "total_tokens": 40267104} +{"current_steps": 40035, "total_steps": 42420, "loss": 0.2027, "lr": 4.803178237178546e-07, "epoch": 18.875530410183874, "percentage": 94.38, "elapsed_time": "2:24:48", "remaining_time": "0:08:37", "throughput": 4634.9, "total_tokens": 40272576} +{"current_steps": 40040, "total_steps": 42420, "loss": 0.109, "lr": 4.783133029865855e-07, "epoch": 18.877887788778878, "percentage": 94.39, "elapsed_time": "2:24:49", "remaining_time": "0:08:36", "throughput": 4634.91, "total_tokens": 40277088} +{"current_steps": 40045, "total_steps": 42420, "loss": 0.0756, "lr": 4.763129333605176e-07, "epoch": 18.88024516737388, "percentage": 94.4, "elapsed_time": "2:24:50", "remaining_time": "0:08:35", "throughput": 4634.93, "total_tokens": 40282112} +{"current_steps": 40050, "total_steps": 42420, "loss": 0.1001, "lr": 4.743167151782857e-07, "epoch": 18.882602545968883, "percentage": 94.41, "elapsed_time": "2:24:51", "remaining_time": "0:08:34", "throughput": 4634.95, "total_tokens": 40286528} +{"current_steps": 40055, "total_steps": 42420, "loss": 0.1566, "lr": 4.7232464877780824e-07, "epoch": 18.884959924563884, "percentage": 94.42, "elapsed_time": "2:24:53", "remaining_time": "0:08:33", "throughput": 4634.99, "total_tokens": 40292384} +{"current_steps": 40060, "total_steps": 42420, "loss": 0.1333, "lr": 4.703367344963128e-07, "epoch": 18.887317303158888, "percentage": 94.44, "elapsed_time": "2:24:54", "remaining_time": "0:08:32", "throughput": 4635.03, "total_tokens": 40298720} +{"current_steps": 40065, "total_steps": 42420, "loss": 0.0778, "lr": 4.683529726703162e-07, "epoch": 18.88967468175389, "percentage": 94.45, "elapsed_time": "2:24:55", "remaining_time": "0:08:31", "throughput": 4635.02, "total_tokens": 40302592} +{"current_steps": 40070, "total_steps": 42420, "loss": 0.1347, "lr": 4.6637336363563876e-07, "epoch": 18.892032060348892, "percentage": 94.46, "elapsed_time": "2:24:56", "remaining_time": "0:08:30", "throughput": 4635.07, "total_tokens": 40309536} +{"current_steps": 40075, "total_steps": 42420, "loss": 0.269, "lr": 4.64397907727393e-07, "epoch": 18.894389438943893, "percentage": 94.47, "elapsed_time": "2:24:57", "remaining_time": "0:08:28", "throughput": 4635.09, "total_tokens": 40314560} +{"current_steps": 40080, "total_steps": 42420, "loss": 0.1538, "lr": 4.6242660527999194e-07, "epoch": 18.896746817538897, "percentage": 94.48, "elapsed_time": "2:24:59", "remaining_time": "0:08:27", "throughput": 4635.12, "total_tokens": 40321152} +{"current_steps": 40085, "total_steps": 42420, "loss": 0.1146, "lr": 4.6045945662714083e-07, "epoch": 18.899104196133898, "percentage": 94.5, "elapsed_time": "2:25:00", "remaining_time": "0:08:26", "throughput": 4635.12, "total_tokens": 40325920} +{"current_steps": 40090, "total_steps": 42420, "loss": 0.0472, "lr": 4.5849646210184837e-07, "epoch": 18.901461574728902, "percentage": 94.51, "elapsed_time": "2:25:01", "remaining_time": "0:08:25", "throughput": 4635.14, "total_tokens": 40331072} +{"current_steps": 40095, "total_steps": 42420, "loss": 0.0106, "lr": 4.565376220364126e-07, "epoch": 18.903818953323903, "percentage": 94.52, "elapsed_time": "2:25:02", "remaining_time": "0:08:24", "throughput": 4635.13, "total_tokens": 40335680} +{"current_steps": 40100, "total_steps": 42420, "loss": 0.1069, "lr": 4.5458293676243214e-07, "epoch": 18.906176331918907, "percentage": 94.53, "elapsed_time": "2:25:03", "remaining_time": "0:08:23", "throughput": 4635.16, "total_tokens": 40340384} +{"current_steps": 40105, "total_steps": 42420, "loss": 0.0789, "lr": 4.5263240661080344e-07, "epoch": 18.908533710513908, "percentage": 94.54, "elapsed_time": "2:25:04", "remaining_time": "0:08:22", "throughput": 4635.21, "total_tokens": 40345888} +{"current_steps": 40110, "total_steps": 42420, "loss": 0.0794, "lr": 4.5068603191171786e-07, "epoch": 18.91089108910891, "percentage": 94.55, "elapsed_time": "2:25:05", "remaining_time": "0:08:21", "throughput": 4635.2, "total_tokens": 40350144} +{"current_steps": 40115, "total_steps": 42420, "loss": 0.1953, "lr": 4.4874381299466475e-07, "epoch": 18.913248467703912, "percentage": 94.57, "elapsed_time": "2:25:06", "remaining_time": "0:08:20", "throughput": 4635.22, "total_tokens": 40354912} +{"current_steps": 40120, "total_steps": 42420, "loss": 0.0445, "lr": 4.468057501884254e-07, "epoch": 18.915605846298917, "percentage": 94.58, "elapsed_time": "2:25:07", "remaining_time": "0:08:19", "throughput": 4635.28, "total_tokens": 40362400} +{"current_steps": 40125, "total_steps": 42420, "loss": 0.0599, "lr": 4.448718438210819e-07, "epoch": 18.917963224893917, "percentage": 94.59, "elapsed_time": "2:25:09", "remaining_time": "0:08:18", "throughput": 4635.32, "total_tokens": 40369760} +{"current_steps": 40130, "total_steps": 42420, "loss": 0.0339, "lr": 4.4294209422001966e-07, "epoch": 18.92032060348892, "percentage": 94.6, "elapsed_time": "2:25:10", "remaining_time": "0:08:17", "throughput": 4635.35, "total_tokens": 40374624} +{"current_steps": 40135, "total_steps": 42420, "loss": 0.0931, "lr": 4.4101650171189946e-07, "epoch": 18.922677982083922, "percentage": 94.61, "elapsed_time": "2:25:11", "remaining_time": "0:08:15", "throughput": 4635.39, "total_tokens": 40380608} +{"current_steps": 40140, "total_steps": 42420, "loss": 0.0337, "lr": 4.390950666226995e-07, "epoch": 18.925035360678926, "percentage": 94.63, "elapsed_time": "2:25:12", "remaining_time": "0:08:14", "throughput": 4635.4, "total_tokens": 40384992} +{"current_steps": 40145, "total_steps": 42420, "loss": 0.0712, "lr": 4.371777892776846e-07, "epoch": 18.927392739273927, "percentage": 94.64, "elapsed_time": "2:25:13", "remaining_time": "0:08:13", "throughput": 4635.42, "total_tokens": 40390464} +{"current_steps": 40150, "total_steps": 42420, "loss": 0.1516, "lr": 4.352646700014146e-07, "epoch": 18.92975011786893, "percentage": 94.65, "elapsed_time": "2:25:14", "remaining_time": "0:08:12", "throughput": 4635.45, "total_tokens": 40396192} +{"current_steps": 40155, "total_steps": 42420, "loss": 0.039, "lr": 4.3335570911775257e-07, "epoch": 18.93210749646393, "percentage": 94.66, "elapsed_time": "2:25:15", "remaining_time": "0:08:11", "throughput": 4635.47, "total_tokens": 40401216} +{"current_steps": 40160, "total_steps": 42420, "loss": 0.0622, "lr": 4.314509069498512e-07, "epoch": 18.934464875058936, "percentage": 94.67, "elapsed_time": "2:25:16", "remaining_time": "0:08:10", "throughput": 4635.49, "total_tokens": 40406432} +{"current_steps": 40165, "total_steps": 42420, "loss": 0.2014, "lr": 4.2955026382016097e-07, "epoch": 18.936822253653936, "percentage": 94.68, "elapsed_time": "2:25:17", "remaining_time": "0:08:09", "throughput": 4635.53, "total_tokens": 40411776} +{"current_steps": 40170, "total_steps": 42420, "loss": 0.0628, "lr": 4.276537800504299e-07, "epoch": 18.93917963224894, "percentage": 94.7, "elapsed_time": "2:25:18", "remaining_time": "0:08:08", "throughput": 4635.53, "total_tokens": 40415808} +{"current_steps": 40175, "total_steps": 42420, "loss": 0.1504, "lr": 4.2576145596169845e-07, "epoch": 18.94153701084394, "percentage": 94.71, "elapsed_time": "2:25:19", "remaining_time": "0:08:07", "throughput": 4635.54, "total_tokens": 40420864} +{"current_steps": 40180, "total_steps": 42420, "loss": 0.0819, "lr": 4.238732918743049e-07, "epoch": 18.943894389438945, "percentage": 94.72, "elapsed_time": "2:25:20", "remaining_time": "0:08:06", "throughput": 4635.56, "total_tokens": 40425152} +{"current_steps": 40185, "total_steps": 42420, "loss": 0.1008, "lr": 4.219892881078824e-07, "epoch": 18.946251768033946, "percentage": 94.73, "elapsed_time": "2:25:21", "remaining_time": "0:08:05", "throughput": 4635.56, "total_tokens": 40429600} +{"current_steps": 40190, "total_steps": 42420, "loss": 0.0987, "lr": 4.201094449813647e-07, "epoch": 18.94860914662895, "percentage": 94.74, "elapsed_time": "2:25:22", "remaining_time": "0:08:03", "throughput": 4635.58, "total_tokens": 40434272} +{"current_steps": 40195, "total_steps": 42420, "loss": 0.1234, "lr": 4.182337628129751e-07, "epoch": 18.95096652522395, "percentage": 94.75, "elapsed_time": "2:25:23", "remaining_time": "0:08:02", "throughput": 4635.61, "total_tokens": 40438784} +{"current_steps": 40200, "total_steps": 42420, "loss": 0.0894, "lr": 4.163622419202345e-07, "epoch": 18.953323903818955, "percentage": 94.77, "elapsed_time": "2:25:24", "remaining_time": "0:08:01", "throughput": 4635.62, "total_tokens": 40442816} +{"current_steps": 40205, "total_steps": 42420, "loss": 0.052, "lr": 4.144948826199618e-07, "epoch": 18.955681282413956, "percentage": 94.78, "elapsed_time": "2:25:25", "remaining_time": "0:08:00", "throughput": 4635.65, "total_tokens": 40448384} +{"current_steps": 40210, "total_steps": 42420, "loss": 0.2825, "lr": 4.1263168522826525e-07, "epoch": 18.95803866100896, "percentage": 94.79, "elapsed_time": "2:25:26", "remaining_time": "0:07:59", "throughput": 4635.69, "total_tokens": 40453888} +{"current_steps": 40215, "total_steps": 42420, "loss": 0.1606, "lr": 4.1077265006055634e-07, "epoch": 18.96039603960396, "percentage": 94.8, "elapsed_time": "2:25:27", "remaining_time": "0:07:58", "throughput": 4635.7, "total_tokens": 40458560} +{"current_steps": 40220, "total_steps": 42420, "loss": 0.1368, "lr": 4.0891777743153334e-07, "epoch": 18.962753418198965, "percentage": 94.81, "elapsed_time": "2:25:28", "remaining_time": "0:07:57", "throughput": 4635.74, "total_tokens": 40463520} +{"current_steps": 40225, "total_steps": 42420, "loss": 0.0841, "lr": 4.0706706765520074e-07, "epoch": 18.965110796793965, "percentage": 94.83, "elapsed_time": "2:25:29", "remaining_time": "0:07:56", "throughput": 4635.78, "total_tokens": 40469408} +{"current_steps": 40230, "total_steps": 42420, "loss": 0.0218, "lr": 4.0522052104484675e-07, "epoch": 18.967468175388966, "percentage": 94.84, "elapsed_time": "2:25:30", "remaining_time": "0:07:55", "throughput": 4635.79, "total_tokens": 40474240} +{"current_steps": 40235, "total_steps": 42420, "loss": 0.0552, "lr": 4.033781379130658e-07, "epoch": 18.96982555398397, "percentage": 94.85, "elapsed_time": "2:25:31", "remaining_time": "0:07:54", "throughput": 4635.81, "total_tokens": 40479424} +{"current_steps": 40240, "total_steps": 42420, "loss": 0.1793, "lr": 4.015399185717389e-07, "epoch": 18.97218293257897, "percentage": 94.86, "elapsed_time": "2:25:32", "remaining_time": "0:07:53", "throughput": 4635.83, "total_tokens": 40483808} +{"current_steps": 40245, "total_steps": 42420, "loss": 0.1387, "lr": 3.9970586333204496e-07, "epoch": 18.974540311173975, "percentage": 94.87, "elapsed_time": "2:25:33", "remaining_time": "0:07:52", "throughput": 4635.83, "total_tokens": 40488352} +{"current_steps": 40250, "total_steps": 42420, "loss": 0.0772, "lr": 3.9787597250445784e-07, "epoch": 18.976897689768975, "percentage": 94.88, "elapsed_time": "2:25:34", "remaining_time": "0:07:50", "throughput": 4635.85, "total_tokens": 40493312} +{"current_steps": 40255, "total_steps": 42420, "loss": 0.0801, "lr": 3.960502463987492e-07, "epoch": 18.97925506836398, "percentage": 94.9, "elapsed_time": "2:25:35", "remaining_time": "0:07:49", "throughput": 4635.85, "total_tokens": 40498112} +{"current_steps": 40260, "total_steps": 42420, "loss": 0.0248, "lr": 3.942286853239857e-07, "epoch": 18.98161244695898, "percentage": 94.91, "elapsed_time": "2:25:36", "remaining_time": "0:07:48", "throughput": 4635.87, "total_tokens": 40502784} +{"current_steps": 40265, "total_steps": 42420, "loss": 0.1995, "lr": 3.924112895885207e-07, "epoch": 18.983969825553984, "percentage": 94.92, "elapsed_time": "2:25:37", "remaining_time": "0:07:47", "throughput": 4635.89, "total_tokens": 40507488} +{"current_steps": 40270, "total_steps": 42420, "loss": 0.1477, "lr": 3.905980595000108e-07, "epoch": 18.986327204148985, "percentage": 94.93, "elapsed_time": "2:25:38", "remaining_time": "0:07:46", "throughput": 4635.9, "total_tokens": 40511744} +{"current_steps": 40275, "total_steps": 42420, "loss": 0.0946, "lr": 3.88788995365405e-07, "epoch": 18.98868458274399, "percentage": 94.94, "elapsed_time": "2:25:39", "remaining_time": "0:07:45", "throughput": 4635.94, "total_tokens": 40517600} +{"current_steps": 40280, "total_steps": 42420, "loss": 0.0613, "lr": 3.869840974909472e-07, "epoch": 18.99104196133899, "percentage": 94.96, "elapsed_time": "2:25:40", "remaining_time": "0:07:44", "throughput": 4635.95, "total_tokens": 40521856} +{"current_steps": 40285, "total_steps": 42420, "loss": 0.076, "lr": 3.851833661821791e-07, "epoch": 18.993399339933994, "percentage": 94.97, "elapsed_time": "2:25:41", "remaining_time": "0:07:43", "throughput": 4635.98, "total_tokens": 40527104} +{"current_steps": 40290, "total_steps": 42420, "loss": 0.0429, "lr": 3.8338680174392906e-07, "epoch": 18.995756718528995, "percentage": 94.98, "elapsed_time": "2:25:43", "remaining_time": "0:07:42", "throughput": 4636.03, "total_tokens": 40532960} +{"current_steps": 40295, "total_steps": 42420, "loss": 0.0632, "lr": 3.815944044803288e-07, "epoch": 18.998114097124, "percentage": 94.99, "elapsed_time": "2:25:44", "remaining_time": "0:07:41", "throughput": 4636.07, "total_tokens": 40538112} +{"current_steps": 40299, "total_steps": 42420, "eval_loss": 0.15307621657848358, "epoch": 19.0, "percentage": 95.0, "elapsed_time": "2:26:00", "remaining_time": "0:07:41", "throughput": 4627.95, "total_tokens": 40541312} +{"current_steps": 40300, "total_steps": 42420, "loss": 0.1769, "lr": 3.7980617469479953e-07, "epoch": 19.000471475719, "percentage": 95.0, "elapsed_time": "2:26:01", "remaining_time": "0:07:40", "throughput": 4627.07, "total_tokens": 40542240} +{"current_steps": 40305, "total_steps": 42420, "loss": 0.1543, "lr": 3.780221126900574e-07, "epoch": 19.002828854314004, "percentage": 95.01, "elapsed_time": "2:26:02", "remaining_time": "0:07:39", "throughput": 4627.04, "total_tokens": 40546560} +{"current_steps": 40310, "total_steps": 42420, "loss": 0.0701, "lr": 3.762422187681136e-07, "epoch": 19.005186232909004, "percentage": 95.03, "elapsed_time": "2:26:04", "remaining_time": "0:07:38", "throughput": 4627.06, "total_tokens": 40551680}